{
  "claude-opus-4": {
    "name": "Claude Opus 4",
    "provider": "Anthropic",
    "releaseDate": "2025-05-14",
    "modelCard": "https://www.anthropic.com/news/claude-4",
    "benchmarks": {
      "arena_text": {
        "score": 1412,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1400,
        "source": "https://lmarena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 70.7,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "aa_intelligence": {
        "score": 33,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "aa_math_index": {
        "score": 36.3,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "gpqa_diamond": {
        "score": 70.1,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "hle": {
        "score": 5.9,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "scicode": {
        "score": 40.9,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "aa_lcr": {
        "score": 36,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "ifbench": {
        "score": 43.3,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "mmlu_pro": {
        "score": 86,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "livecodebench": {
        "score": 54.2,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      },
      "aime_25": {
        "score": 36.3,
        "source": "https://artificialanalysis.ai/models/claude-4-opus"
      }
    },
    "api": {
      "inputPer1M": 18.75,
      "outputPer1M": 75,
      "source": "https://artificialanalysis.ai/models/claude-4-opus",
      "provider": "Anthropic",
      "tokensPerSecAPI": 34.4
    },
    "local": null,
    "subscriptions": {
      "claude_max_20x": {
        "name": "Claude Max 20×",
        "monthlyPrice": 200,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Max 20×. Multi-flip method (3 weekly-% crossings, first discarded). First real Opus-on-Max measurement; prior Opus rows used the Sonnet-derived 203M as a placeholder. Source: claude_measurement_20260424_094106.json.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_094106.json",
        "tokensPerWeek": 248361650,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_094106.json",
        "measurementTaskRuns": null,
        "measurementDelta5h": 18,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)"
      },
      "claude_pro": {
        "name": "Claude Pro",
        "monthlyPrice": 20,
        "confidence": "medium",
        "tokensPerWeek": 15595812,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_161653.json",
        "measurementTaskRuns": 20,
        "measurementDelta5h": 100,
        "measurementDeltaWeekly": 8,
        "measurement_method": "coarse single-delta (20-parallel batch moved weekly 0%→8% in one step; multi-flip data invalidated by all 3 crossings occurring in same batch — known script limitation with large PARALLEL on small-quota plans)",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Pro. First real Pro measurement (Opus specifically). Coarse single-delta method: 1.25M tokens consumed 8% of weekly quota, extrapolated to ~15.6M tok/wk. Session meter went 0→100% confirming Pro has much tighter quota than Max. Sonnet-on-Pro measurement pending (weekly will reset Apr 28). Source: claude_measurement_20260424_161653.json.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_161653.json"
      },
      "claude_max_5x": {
        "name": "Claude Max 5×",
        "monthlyPrice": 100,
        "tokensPerWeek": 62090413,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "claude_max_20x",
          "base_tokens_per_week": 248361650,
          "base_measured_on_model": "Opus 4.7",
          "multiplier": 0.25,
          "multiplier_label": "5/20",
          "source": "https://www.anthropic.com/pricing"
        },
        "confidence": "low",
        "notes": "Anthropic's Max plans are tiered as 5× and 20× Pro. We measure Max 20×; Max 5× is derived as 5/20 of the measured Max 20× number for the same model. Not directly measured.",
        "source": "https://www.anthropic.com/pricing"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "claude-sonnet-4": {
    "name": "Claude Sonnet 4",
    "provider": "Anthropic",
    "releaseDate": "2025-05-14",
    "modelCard": "https://www.anthropic.com/news/claude-4",
    "benchmarks": {
      "arena_text": {
        "score": 1389,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1370,
        "source": "https://lmarena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 56.4,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "livebench": {
        "score": 72,
        "source": "https://livebench.ai/"
      },
      "aa_intelligence": {
        "score": 33,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "aa_coding_index": {
        "score": 30.6,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "aa_math_index": {
        "score": 38,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "gpqa_diamond": {
        "score": 68.3,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "hle": {
        "score": 4,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "scicode": {
        "score": 37.3,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "aa_lcr": {
        "score": 44.3,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "tau2_bench": {
        "score": 52.3,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "terminalbench_hard": {
        "score": 27.3,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "ifbench": {
        "score": 45.4,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "mmlu_pro": {
        "score": 83.7,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "livecodebench": {
        "score": 44.9,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      },
      "aime_25": {
        "score": 38,
        "source": "https://artificialanalysis.ai/models/claude-4-sonnet"
      }
    },
    "api": {
      "inputPer1M": 3.75,
      "outputPer1M": 15,
      "source": "https://artificialanalysis.ai/models/claude-4-sonnet",
      "provider": "Anthropic",
      "tokensPerSecAPI": 46.5
    },
    "local": null,
    "subscriptions": {
      "claude_pro": {
        "name": "Claude Pro",
        "monthlyPrice": 20,
        "confidence": "high",
        "notes": "Measured 2026-04-24 21:39 via claude-code --model sonnet on Sonnet 4.6 on Claude Pro. Multi-flip method, 3 flips in batches 4/7/11, per-1% deltas 168K + 224K (mean 196K/1%). 34% of 5h session consumed, 3% of weekly consumed. Clean run with no same-batch flip pileup. First proper Pro + Sonnet measurement. Source: claude_measurement_20260424_213954.json",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_213954.json",
        "tokensPerWeek": 19639400,
        "estimated": false,
        "measuredOnModel": "Sonnet 4.6",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_213954.json",
        "measurementTaskRuns": 33,
        "measurementDelta5h": 34,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded)"
      },
      "claude_max_20x": {
        "name": "Claude Max 20×",
        "monthlyPrice": 200,
        "confidence": "high",
        "tokensPerWeek": 387831650,
        "estimated": false,
        "measuredOnModel": "Sonnet 4.6",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_110655.json",
        "measurementTaskRuns": null,
        "measurementDelta5h": 16,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)",
        "notes": "Measured 2026-04-24 via claude-code --model sonnet on Sonnet 4.6 on Claude Max 20×. Multi-flip method (3 weekly-% crossings, first discarded). Source: claude_measurement_20260424_110655.json. Supersedes the Apr 16 Sonnet 4.5 single-flip run which gave 203M.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_110655.json"
      },
      "claude_max_5x": {
        "name": "Claude Max 5×",
        "monthlyPrice": 100,
        "tokensPerWeek": 96957913,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "claude_max_20x",
          "base_tokens_per_week": 387831650,
          "base_measured_on_model": "Sonnet 4.6",
          "multiplier": 0.25,
          "multiplier_label": "5/20",
          "source": "https://www.anthropic.com/pricing"
        },
        "confidence": "low",
        "notes": "Anthropic's Max plans are tiered as 5× and 20× Pro. We measure Max 20×; Max 5× is derived as 5/20 of the measured Max 20× number for the same model. Not directly measured.",
        "source": "https://www.anthropic.com/pricing"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "deepseek-v3": {
    "name": "DeepSeek V3.1",
    "provider": "DeepSeek",
    "releaseDate": "2025-08-01",
    "modelCard": "https://github.com/deepseek-ai/DeepSeek-V3",
    "benchmarks": {
      "arena_text": {
        "score": 1395,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1350,
        "source": "https://lmarena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 48.4,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "aa_intelligence": {
        "score": 16.5,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "aa_coding_index": {
        "score": 16.4,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "aa_math_index": {
        "score": 26,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "gpqa_diamond": {
        "score": 55.7,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "hle": {
        "score": 3.6,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "scicode": {
        "score": 35.4,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "aa_lcr": {
        "score": 29,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "tau2_bench": {
        "score": 22.8,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "terminalbench_hard": {
        "score": 6.8,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "ifbench": {
        "score": 34.8,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "mmlu_pro": {
        "score": 75.2,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "livecodebench": {
        "score": 35.9,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "aime_25": {
        "score": 26,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "gdpval_aa": {
        "score": 418.8,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "aa_omniscience": {
        "score": -41.4,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      },
      "critpt": {
        "score": 0,
        "source": "https://artificialanalysis.ai/models/deepseek-v3"
      }
    },
    "api": {
      "inputPer1M": 0.4,
      "outputPer1M": 0.89,
      "source": "https://artificialanalysis.ai/models/deepseek-v3",
      "provider": "DeepSeek",
      "tokensPerSecAPI": null
    },
    "local": {
      "rtx_4090": {
        "tokensPerSec": 15,
        "quantization": "Q4_K_M",
        "vramRequired": 80,
        "notes": "Requires multi-GPU or heavy offloading (671B params)",
        "source": "https://github.com/ggerganov/llama.cpp/discussions"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "gemini-2.5-pro": {
    "name": "Gemini 2.5 Pro",
    "provider": "Google",
    "releaseDate": "2025-03-25",
    "modelCard": "https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/",
    "benchmarks": {
      "arena_text": {
        "score": 1448,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1203,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 72.9,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "aa_intelligence": {
        "score": 34.6,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "aa_coding_index": {
        "score": 31.9,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "aa_math_index": {
        "score": 87.7,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "gpqa_diamond": {
        "score": 84.4,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "hle": {
        "score": 21.1,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "scicode": {
        "score": 42.8,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "aa_lcr": {
        "score": 66,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "tau2_bench": {
        "score": 54.1,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "terminalbench_hard": {
        "score": 26.5,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "ifbench": {
        "score": 48.7,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "mmlu_pro": {
        "score": 86.2,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "livecodebench": {
        "score": 80.1,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      },
      "aime_25": {
        "score": 87.7,
        "source": "https://artificialanalysis.ai/models/gemini-2-5-pro"
      }
    },
    "api": {
      "inputPer1M": 1.25,
      "outputPer1M": 10,
      "source": "https://artificialanalysis.ai/models/gemini-2-5-pro",
      "provider": "Google",
      "tokensPerSecAPI": 134.5
    },
    "local": null,
    "subscriptions": {
      "gemini_advanced": {
        "name": "Gemini Advanced",
        "monthlyPrice": 20,
        "confidence": "low",
        "notes": "Estimated 100 prompts/day. Per OpenRouter 2025 study, avg turn is 6,400 tokens (6k in / 400 out).",
        "source": "https://a16z.com/state-of-ai/",
        "tokensPerWeek": 4480000,
        "estimated": true
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "glm-4.7-flash": {
    "name": "GLM-4.7 Flash",
    "provider": "Z.ai",
    "releaseDate": "2026-01-19",
    "modelCard": "https://huggingface.co/zai-org/GLM-4.7-Flash",
    "benchmarks": {
      "gpqa_diamond": {
        "score": 58.1,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      },
      "swe_bench": {
        "score": 59.2,
        "source": "https://huggingface.co/zai-org/GLM-4.7-Flash"
      },
      "aa_intelligence": {
        "score": 30.1,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      },
      "arena_text": {
        "score": 1368,
        "source": "https://arena.ai/leaderboard/text"
      },
      "aa_coding_index": {
        "score": 25.9,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      },
      "hle": {
        "score": 7.1,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      },
      "scicode": {
        "score": 33.7,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      },
      "aa_lcr": {
        "score": 35,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      },
      "tau2_bench": {
        "score": 98.8,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      },
      "terminalbench_hard": {
        "score": 22,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      },
      "ifbench": {
        "score": 60.8,
        "source": "https://artificialanalysis.ai/models/glm-4-7-flash"
      }
    },
    "api": {
      "inputPer1M": 0.07,
      "outputPer1M": 0.4,
      "source": "https://artificialanalysis.ai/models/glm-4-7-flash",
      "provider": "Z.ai",
      "tokensPerSecAPI": 90.4
    },
    "local": {
      "mac_m4_48gb": {
        "tokensPerSec": 16,
        "quantization": "unknown",
        "vramRequired": 48,
        "notes": "User test on MacBook M4 48GB.",
        "source": "user-reported"
      },
      "rtx_500_ada_laptop": {
        "tokensPerSec": 2.3,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 2,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 2 msgs. DB variant: glm-4.7-flash:latest"
      },
      "amd_780m_igpu": {
        "tokensPerSec": 0.9,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 1,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 1 msgs. DB variant: glm-4.7-flash:latest"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "gpt-oss-120b": {
    "name": "GPT-OSS 120B",
    "provider": "OpenAI",
    "releaseDate": "2025-08-05",
    "modelCard": "https://openai.com/index/introducing-gpt-oss/",
    "benchmarks": {
      "gpqa_diamond": {
        "score": 78.2,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "mmlu": {
        "score": 90,
        "source": "https://www.amd.com/en/blogs/2026/amd-ryzen-ai-max-ai-pcs-deliver-exceptional-intelligence.html"
      },
      "arena_text": {
        "score": 1353,
        "source": "https://arena.ai/leaderboard/text"
      },
      "aa_intelligence": {
        "score": 33.3,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "aa_coding_index": {
        "score": 28.6,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "aa_math_index": {
        "score": 93.4,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "hle": {
        "score": 18.5,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "scicode": {
        "score": 38.9,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "aa_lcr": {
        "score": 50.7,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "tau2_bench": {
        "score": 65.8,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "terminalbench_hard": {
        "score": 23.5,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "ifbench": {
        "score": 69,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "mmlu_pro": {
        "score": 80.8,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "livecodebench": {
        "score": 87.8,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "aime_25": {
        "score": 93.4,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "gdpval_aa": {
        "score": 946.6,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "aa_omniscience": {
        "score": -50,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      },
      "critpt": {
        "score": 1.1,
        "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
      }
    },
    "api": {
      "inputPer1M": 0.15,
      "outputPer1M": 0.6,
      "tokensPerSecAPI": 216.4,
      "source": "https://artificialanalysis.ai/models/gpt-oss-120b"
    },
    "local": {
      "framework_desktop_128gb": {
        "tokensPerSec": 33,
        "quantization": "MXFP4 (GGUF)",
        "notes": "Community benchmark on Framework Desktop; Vulkan backend, LM Studio/llama.cpp.",
        "source": "https://community.frame.work/t/tracking-will-the-ai-max-395-128gb-be-able-to-run-gpt-oss-120b/73280"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "gpt-4o": {
    "name": "GPT-4o",
    "provider": "OpenAI",
    "releaseDate": "2024-05-13",
    "modelCard": "https://openai.com/index/hello-gpt-4o/",
    "benchmarks": {
      "arena_text": {
        "score": 1443,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1310,
        "source": "https://lmarena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 23.1,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "mmlu": {
        "score": 88.7,
        "source": "https://openai.com/index/hello-gpt-4o/"
      },
      "aa_intelligence": {
        "score": 17.3,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "aa_coding_index": {
        "score": 16.7,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "aa_math_index": {
        "score": 6,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "gpqa_diamond": {
        "score": 54.3,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "hle": {
        "score": 3.3,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "scicode": {
        "score": 33.3,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "aa_lcr": {
        "score": 0,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "tau2_bench": {
        "score": 25.1,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "terminalbench_hard": {
        "score": 8.3,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "ifbench": {
        "score": 34.3,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "mmlu_pro": {
        "score": 74.8,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "livecodebench": {
        "score": 30.9,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "aime_25": {
        "score": 6,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "gdpval_aa": {
        "score": 356.3,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "aa_omniscience": {
        "score": -10.8,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      },
      "critpt": {
        "score": 0,
        "source": "https://artificialanalysis.ai/models/gpt-4o"
      }
    },
    "api": {
      "inputPer1M": 2.5,
      "outputPer1M": 10,
      "source": "https://artificialanalysis.ai/models/gpt-4o",
      "provider": "OpenAI",
      "tokensPerSecAPI": 158
    },
    "local": null,
    "subscriptions": {
      "chatgpt_plus": {
        "name": "ChatGPT Plus",
        "monthlyPrice": 20,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via codex-cli on gpt-5.4 xhigh on Plus. Multi-flip method. Note: this row shows ChatGPT Plus capacity on GPT-4o, but actual measurement was on GPT-5.4 — older Plus-tier models share the same weekly quota meter so the tokens/week figure applies, but quality is different.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/codex_measurement_20260424_110957.json",
        "tokensPerWeek": 190199950,
        "estimated": false,
        "measuredOnModel": "gpt-5.4 (reasoning xhigh, summaries auto)",
        "measurementFiles": 7,
        "measurementBestFile": "codex_measurement_20260424_110957.json",
        "measurementTaskRuns": 40,
        "measurementDelta5h": 21,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)"
      },
      "chatgpt_business": {
        "name": "ChatGPT Business",
        "monthlyPrice": 30,
        "confidence": "medium",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/codex_measurement_20260424_230604.json",
        "notes": "Business quota measured on GPT-5.4 (not directly on GPT-4o), but weekly quota is per-account shared across models. Measured 2026-04-24 23:06 via codex-cli on gpt-5.4 xhigh on Business. Clean multi-flip: 3 flips in 3 separate batches (batches 1/3/4). Per-1% usable deltas: 356,526 and 567,666 — mean 462,096/1%. The 59% spread between the two deltas suggests the 3-flip sample is noisier than ideal; a 5-flip measurement would tighten the estimate. Coarse 3%-span gives 36.8M. Supersedes the contaminated 22:28 measurement (27M, multi-flip broken) and the Apr 15 single-flip measurement (60M, likely high-end outlier). Cache hit rate 80%.",
        "tokensPerWeek": 46209600,
        "estimated": false,
        "measuredOnModel": "gpt-5.4 (reasoning xhigh, summaries auto)",
        "measurementFiles": 1,
        "measurementBestFile": "codex_measurement_20260424_230604.json",
        "measurementTaskRuns": 4,
        "measurementDelta5h": 22,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded)"
      },
      "chatgpt_pro_100": {
        "name": "ChatGPT Pro $100",
        "monthlyPrice": 100,
        "tokensPerWeek": 950999750,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 190199950,
          "base_measured_on_model": "gpt-5.4 (reasoning xhigh, summaries auto)",
          "multiplier": 5,
          "multiplier_label": "5×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $100 plan launched Apr 2026. OpenAI states 5× higher weekly limits than Plus (10× Codex usage through May 31, 2026 as a launch promo). We have not measured Pro $100 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if OpenAI's \"5×\" applies to a dimension other than tokens (e.g. rate-limit slots, message count). Help us verify by running scripts/measure-codex-quota.sh on a Pro $100 account.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      },
      "chatgpt_pro_200": {
        "name": "ChatGPT Pro $200",
        "monthlyPrice": 200,
        "tokensPerWeek": 3803999000,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 190199950,
          "base_measured_on_model": "gpt-5.4 (reasoning xhigh, summaries auto)",
          "multiplier": 20,
          "multiplier_label": "20×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $200 plan. OpenAI states 20× higher weekly limits than Plus (with a 25× Codex 5-hour bonus through May 31, 2026). We have not measured Pro $200 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if \"20×\" applies to a different dimension than tokens.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "llama-3.1-70b": {
    "name": "Llama 3.1 70B",
    "provider": "Meta",
    "releaseDate": "2024-07-23",
    "modelCard": "https://ai.meta.com/blog/meta-llama-3-1/",
    "benchmarks": {
      "arena_text": {
        "score": 1293,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1180,
        "source": "https://lmarena.ai/leaderboard/code"
      },
      "mmlu": {
        "score": 86,
        "source": "https://ai.meta.com/blog/meta-llama-3-1/"
      },
      "aider_polyglot": {
        "score": 32.5,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "aa_intelligence": {
        "score": 12.5,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "aa_coding_index": {
        "score": 10.9,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "aa_math_index": {
        "score": 4,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "gpqa_diamond": {
        "score": 40.9,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "hle": {
        "score": 4.6,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "scicode": {
        "score": 26.7,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "aa_lcr": {
        "score": 6.3,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "tau2_bench": {
        "score": 15.2,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "terminalbench_hard": {
        "score": 3,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "ifbench": {
        "score": 34.4,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "mmlu_pro": {
        "score": 67.6,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "livecodebench": {
        "score": 23.2,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      },
      "aime_25": {
        "score": 4,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b"
      }
    },
    "api": {
      "inputPer1M": 0.56,
      "outputPer1M": 0.56,
      "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-70b",
      "provider": "Together.ai",
      "tokensPerSecAPI": 31.9
    },
    "local": {
      "rtx_3090": {
        "tokensPerSec": 25,
        "quantization": "Q4_K_M",
        "vramRequired": 40,
        "notes": "Requires 2x RTX 3090 or offloading to RAM",
        "source": "https://github.com/ggerganov/llama.cpp/discussions/4225"
      },
      "rtx_4090": {
        "tokensPerSec": 45,
        "quantization": "Q4_K_M",
        "vramRequired": 40,
        "source": "https://github.com/ggerganov/llama.cpp/discussions"
      },
      "mac_m3_max_128gb": {
        "tokensPerSec": 18,
        "quantization": "Q4_K_M",
        "vramRequired": 45,
        "source": "https://github.com/ggerganov/llama.cpp/discussions/4167"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "llama-3.1-8b": {
    "name": "Llama 3.1 8B",
    "provider": "Meta",
    "releaseDate": "2024-07-23",
    "modelCard": "https://ai.meta.com/blog/meta-llama-3-1/",
    "benchmarks": {
      "arena_text": {
        "score": 1211,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1080,
        "source": "https://lmarena.ai/leaderboard/code"
      },
      "mmlu": {
        "score": 69.4,
        "source": "https://ai.meta.com/blog/meta-llama-3-1/"
      },
      "aa_intelligence": {
        "score": 11.8,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "aa_coding_index": {
        "score": 4.9,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "aa_math_index": {
        "score": 4.3,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "gpqa_diamond": {
        "score": 25.9,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "hle": {
        "score": 5.1,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "scicode": {
        "score": 13.2,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "aa_lcr": {
        "score": 15.7,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "tau2_bench": {
        "score": 16.4,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "terminalbench_hard": {
        "score": 0.8,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "ifbench": {
        "score": 28.6,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "mmlu_pro": {
        "score": 47.6,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "livecodebench": {
        "score": 11.6,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      },
      "aime_25": {
        "score": 4.3,
        "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b"
      }
    },
    "api": {
      "inputPer1M": 0.1,
      "outputPer1M": 0.1,
      "source": "https://artificialanalysis.ai/models/llama-3-1-instruct-8b",
      "provider": "Together.ai",
      "tokensPerSecAPI": 203.3
    },
    "local": {
      "rtx_3090": {
        "tokensPerSec": 120,
        "quantization": "Q4_K_M",
        "vramRequired": 6,
        "source": "https://github.com/ggerganov/llama.cpp/discussions/4225"
      },
      "rtx_4090": {
        "tokensPerSec": 216,
        "quantization": "Q4_K_M",
        "vramRequired": 6,
        "source": "https://github.com/ggerganov/llama.cpp/discussions"
      },
      "mac_m3_max_64gb": {
        "tokensPerSec": 85,
        "quantization": "Q4_K_M",
        "vramRequired": 6,
        "source": "https://github.com/ggerganov/llama.cpp/discussions/4167"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "o3": {
    "name": "o3",
    "provider": "OpenAI",
    "releaseDate": "2025-04-16",
    "modelCard": "https://openai.com/o3",
    "benchmarks": {
      "arena_text": {
        "score": 1431,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1420,
        "source": "https://lmarena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 76.9,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "aa_intelligence": {
        "score": 38.4,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "aa_coding_index": {
        "score": 38.4,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "aa_math_index": {
        "score": 88.3,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "gpqa_diamond": {
        "score": 82.7,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "hle": {
        "score": 20,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "scicode": {
        "score": 41,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "aa_lcr": {
        "score": 69.3,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "tau2_bench": {
        "score": 80.7,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "terminalbench_hard": {
        "score": 37.1,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "ifbench": {
        "score": 71.4,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "mmlu_pro": {
        "score": 85.3,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "livecodebench": {
        "score": 80.8,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "aime_25": {
        "score": 88.3,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "gdpval_aa": {
        "score": 755.6,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "aa_omniscience": {
        "score": -15.3,
        "source": "https://artificialanalysis.ai/models/o3"
      },
      "critpt": {
        "score": 1.1,
        "source": "https://artificialanalysis.ai/models/o3"
      }
    },
    "api": {
      "inputPer1M": 2,
      "outputPer1M": 8,
      "source": "https://artificialanalysis.ai/models/o3",
      "provider": "OpenAI",
      "tokensPerSecAPI": 110.3
    },
    "local": null,
    "subscriptions": {},
    "_aa_synced": "2026-05-11"
  },
  "qwen-3-32b": {
    "name": "Qwen QwQ-32B",
    "provider": "Alibaba",
    "releaseDate": "2025-04-29",
    "modelCard": "https://huggingface.co/Qwen/QwQ-32B",
    "benchmarks": {
      "arena_text": {
        "score": 1336,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1200,
        "source": "https://lmarena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 40,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "aa_intelligence": {
        "score": 19.7,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "aa_math_index": {
        "score": 29,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "gpqa_diamond": {
        "score": 59.3,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "hle": {
        "score": 8.2,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "scicode": {
        "score": 35.8,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "aa_lcr": {
        "score": 25,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "ifbench": {
        "score": 38.8,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "mmlu_pro": {
        "score": 76.4,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "livecodebench": {
        "score": 63.1,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      },
      "aime_25": {
        "score": 29,
        "source": "https://artificialanalysis.ai/models/qwq-32b"
      }
    },
    "api": {
      "inputPer1M": 0.66,
      "outputPer1M": 1,
      "source": "https://artificialanalysis.ai/models/qwq-32b",
      "provider": "Together.ai",
      "tokensPerSecAPI": 30.7
    },
    "local": {
      "rtx_3090": {
        "tokensPerSec": 45,
        "quantization": "Q4_K_M",
        "vramRequired": 20,
        "source": "https://github.com/ggerganov/llama.cpp/discussions"
      },
      "rtx_4090": {
        "tokensPerSec": 81,
        "quantization": "Q4_K_M",
        "vramRequired": 20,
        "source": "https://github.com/ggerganov/llama.cpp/discussions"
      },
      "mac_m3_max_64gb": {
        "tokensPerSec": 35,
        "quantization": "Q4_K_M",
        "vramRequired": 22,
        "source": "https://github.com/ggerganov/llama.cpp/discussions/4167"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "claude-opus-4-5": {
    "name": "Claude Opus 4.5",
    "provider": "Anthropic",
    "releaseDate": "2025-11-24",
    "modelCard": "https://www.anthropic.com/news/claude-opus-4-5",
    "benchmarks": {
      "arena_text": {
        "score": 1468,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1467,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 89.4,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "swe_bench": {
        "score": 80.9,
        "source": "https://www.swebench.com/"
      },
      "aa_intelligence": {
        "score": 43.1,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "aa_coding_index": {
        "score": 42.9,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "aa_math_index": {
        "score": 62.7,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "gpqa_diamond": {
        "score": 81,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "hle": {
        "score": 12.9,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "scicode": {
        "score": 47,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "aa_lcr": {
        "score": 65.3,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "tau2_bench": {
        "score": 86.3,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "terminalbench_hard": {
        "score": 40.9,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "ifbench": {
        "score": 43,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "mmlu_pro": {
        "score": 88.9,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "livecodebench": {
        "score": 73.8,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "aime_25": {
        "score": 62.7,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "gdpval_aa": {
        "score": 1424.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "aa_omniscience": {
        "score": -4,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      },
      "critpt": {
        "score": 0.3,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-5"
      }
    },
    "api": {
      "inputPer1M": 6.25,
      "outputPer1M": 25,
      "source": "https://artificialanalysis.ai/models/claude-opus-4-5",
      "provider": "Anthropic",
      "tokensPerSecAPI": 53.6
    },
    "local": null,
    "subscriptions": {
      "claude_max_20x": {
        "name": "Claude Max 20×",
        "monthlyPrice": 200,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Max 20×. Multi-flip method (3 weekly-% crossings, first discarded). First real Opus-on-Max measurement; prior Opus rows used the Sonnet-derived 203M as a placeholder. Source: claude_measurement_20260424_094106.json.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_094106.json",
        "tokensPerWeek": 248361650,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_094106.json",
        "measurementTaskRuns": null,
        "measurementDelta5h": 18,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)"
      },
      "claude_pro": {
        "name": "Claude Pro",
        "monthlyPrice": 20,
        "confidence": "medium",
        "tokensPerWeek": 15595812,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_161653.json",
        "measurementTaskRuns": 20,
        "measurementDelta5h": 100,
        "measurementDeltaWeekly": 8,
        "measurement_method": "coarse single-delta (20-parallel batch moved weekly 0%→8% in one step; multi-flip data invalidated by all 3 crossings occurring in same batch — known script limitation with large PARALLEL on small-quota plans)",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Pro. First real Pro measurement (Opus specifically). Coarse single-delta method: 1.25M tokens consumed 8% of weekly quota, extrapolated to ~15.6M tok/wk. Session meter went 0→100% confirming Pro has much tighter quota than Max. Sonnet-on-Pro measurement pending (weekly will reset Apr 28). Source: claude_measurement_20260424_161653.json.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_161653.json"
      },
      "claude_max_5x": {
        "name": "Claude Max 5×",
        "monthlyPrice": 100,
        "tokensPerWeek": 62090413,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "claude_max_20x",
          "base_tokens_per_week": 248361650,
          "base_measured_on_model": "Opus 4.7",
          "multiplier": 0.25,
          "multiplier_label": "5/20",
          "source": "https://www.anthropic.com/pricing"
        },
        "confidence": "low",
        "notes": "Anthropic's Max plans are tiered as 5× and 20× Pro. We measure Max 20×; Max 5× is derived as 5/20 of the measured Max 20× number for the same model. Not directly measured.",
        "source": "https://www.anthropic.com/pricing"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "gemini-3-pro": {
    "name": "Gemini 3 Pro",
    "provider": "Google",
    "releaseDate": "2025-11-18",
    "modelCard": "https://blog.google/products/gemini/gemini-3-pro/",
    "benchmarks": {
      "arena_text": {
        "score": 1486,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1438,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 75,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "swe_bench": {
        "score": 76.2,
        "source": "https://www.swebench.com/"
      },
      "aa_intelligence": {
        "score": 48.4,
        "source": "https://artificialanalysis.ai/"
      },
      "mmlu_pro": {
        "score": 89.8,
        "source": "https://artificialanalysis.ai/models/gemini-3-pro"
      },
      "livecodebench": {
        "score": 91.7,
        "source": "https://artificialanalysis.ai/models/gemini-3-pro"
      },
      "aime_25": {
        "score": 95.7,
        "source": "https://artificialanalysis.ai/models/gemini-3-pro"
      },
      "gdpval_aa": {
        "score": 1189.7,
        "source": "https://artificialanalysis.ai/models/gemini-3-pro"
      },
      "aa_omniscience": {
        "score": 15.8,
        "source": "https://artificialanalysis.ai/models/gemini-3-pro"
      },
      "critpt": {
        "score": 9.1,
        "source": "https://artificialanalysis.ai/models/gemini-3-pro"
      }
    },
    "api": {
      "inputPer1M": 2,
      "outputPer1M": 12,
      "source": "https://ai.google.dev/gemini-api/docs/pricing",
      "provider": "Google"
    },
    "local": null,
    "subscriptions": {
      "gemini_advanced": {
        "name": "Google AI Ultra",
        "monthlyPrice": 20,
        "confidence": "low",
        "notes": "Priority access. Avg turn 6,400 tokens (OpenRouter 2025 Study).",
        "source": "https://a16z.com/state-of-ai/",
        "tokensPerWeek": 4480000,
        "estimated": true
      }
    }
  },
  "gemini-3-flash": {
    "name": "Gemini 3 Flash",
    "provider": "Google",
    "releaseDate": "2025-12-17",
    "modelCard": "https://blog.google/products/gemini/gemini-3-flash/",
    "benchmarks": {
      "arena_text": {
        "score": 1474,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1437,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 72,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "swe_bench": {
        "score": 78,
        "source": "https://www.swebench.com/"
      },
      "aa_intelligence": {
        "score": 35,
        "source": "https://artificialanalysis.ai/"
      },
      "mmlu_pro": {
        "score": 88.2,
        "source": "https://artificialanalysis.ai/models/gemini-3-flash"
      },
      "livecodebench": {
        "score": 79.7,
        "source": "https://artificialanalysis.ai/models/gemini-3-flash"
      },
      "aime_25": {
        "score": 55.7,
        "source": "https://artificialanalysis.ai/models/gemini-3-flash"
      },
      "gdpval_aa": {
        "score": 1118,
        "source": "https://artificialanalysis.ai/models/gemini-3-flash"
      },
      "aa_omniscience": {
        "score": -3.6,
        "source": "https://artificialanalysis.ai/models/gemini-3-flash"
      },
      "critpt": {
        "score": 1.4,
        "source": "https://artificialanalysis.ai/models/gemini-3-flash"
      }
    },
    "api": {
      "inputPer1M": 0.5,
      "outputPer1M": 3,
      "source": "https://ai.google.dev/gemini-api/docs/pricing",
      "provider": "Google"
    },
    "local": null,
    "subscriptions": null
  },
  "gpt-5-2": {
    "name": "GPT-5.2",
    "provider": "OpenAI",
    "releaseDate": "2025-12-11",
    "modelCard": "https://openai.com/index/introducing-gpt-5-2/",
    "benchmarks": {
      "arena_text": {
        "score": 1437,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1404,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aider_polyglot": {
        "score": 88,
        "source": "https://aider.chat/docs/leaderboards/"
      },
      "swe_bench": {
        "score": 80,
        "source": "https://www.swebench.com/"
      },
      "aa_intelligence": {
        "score": 51.3,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "aa_coding_index": {
        "score": 48.7,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "aa_math_index": {
        "score": 99,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "gpqa_diamond": {
        "score": 90.3,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "hle": {
        "score": 35.4,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "scicode": {
        "score": 52.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "aa_lcr": {
        "score": 72.7,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "tau2_bench": {
        "score": 84.8,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "terminalbench_hard": {
        "score": 47,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "ifbench": {
        "score": 75.4,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "mmlu_pro": {
        "score": 87.4,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "livecodebench": {
        "score": 88.9,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "aime_25": {
        "score": 99,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "gdpval_aa": {
        "score": 1469.3,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "aa_omniscience": {
        "score": -1,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      },
      "critpt": {
        "score": 11.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-2"
      }
    },
    "api": {
      "inputPer1M": 1.75,
      "outputPer1M": 14,
      "source": "https://artificialanalysis.ai/models/gpt-5-2",
      "provider": "OpenAI",
      "tokensPerSecAPI": 80.1
    },
    "local": null,
    "subscriptions": {
      "chatgpt_plus": {
        "name": "ChatGPT Plus",
        "monthlyPrice": 20,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via codex-cli on gpt-5.4 xhigh on Plus. Multi-flip method. Note: this row shows ChatGPT Plus capacity on GPT-5.2, but actual measurement was on GPT-5.4 — older Plus-tier models share the same weekly quota meter so the tokens/week figure applies, but quality is different.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/codex_measurement_20260424_110957.json",
        "tokensPerWeek": 190199950,
        "estimated": false,
        "measuredOnModel": "gpt-5.4 (reasoning xhigh, summaries auto)",
        "measurementFiles": 7,
        "measurementBestFile": "codex_measurement_20260424_110957.json",
        "measurementTaskRuns": 40,
        "measurementDelta5h": 21,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)"
      },
      "chatgpt_business": {
        "name": "ChatGPT Business",
        "monthlyPrice": 30,
        "confidence": "medium",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/codex_measurement_20260424_230604.json",
        "notes": "Business quota measured on GPT-5.4 (not directly on GPT-5.2), but weekly quota is per-account shared across models. Measured 2026-04-24 23:06 via codex-cli on gpt-5.4 xhigh on Business. Clean multi-flip: 3 flips in 3 separate batches (batches 1/3/4). Per-1% usable deltas: 356,526 and 567,666 — mean 462,096/1%. The 59% spread between the two deltas suggests the 3-flip sample is noisier than ideal; a 5-flip measurement would tighten the estimate. Coarse 3%-span gives 36.8M. Supersedes the contaminated 22:28 measurement (27M, multi-flip broken) and the Apr 15 single-flip measurement (60M, likely high-end outlier). Cache hit rate 80%.",
        "tokensPerWeek": 46209600,
        "estimated": false,
        "measuredOnModel": "gpt-5.4 (reasoning xhigh, summaries auto)",
        "measurementFiles": 1,
        "measurementBestFile": "codex_measurement_20260424_230604.json",
        "measurementTaskRuns": 4,
        "measurementDelta5h": 22,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded)"
      },
      "chatgpt_pro_100": {
        "name": "ChatGPT Pro $100",
        "monthlyPrice": 100,
        "tokensPerWeek": 950999750,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 190199950,
          "base_measured_on_model": "gpt-5.4 (reasoning xhigh, summaries auto)",
          "multiplier": 5,
          "multiplier_label": "5×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $100 plan launched Apr 2026. OpenAI states 5× higher weekly limits than Plus (10× Codex usage through May 31, 2026 as a launch promo). We have not measured Pro $100 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if OpenAI's \"5×\" applies to a dimension other than tokens (e.g. rate-limit slots, message count). Help us verify by running scripts/measure-codex-quota.sh on a Pro $100 account.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      },
      "chatgpt_pro_200": {
        "name": "ChatGPT Pro $200",
        "monthlyPrice": 200,
        "tokensPerWeek": 3803999000,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 190199950,
          "base_measured_on_model": "gpt-5.4 (reasoning xhigh, summaries auto)",
          "multiplier": 20,
          "multiplier_label": "20×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $200 plan. OpenAI states 20× higher weekly limits than Plus (with a 25× Codex 5-hour bonus through May 31, 2026). We have not measured Pro $200 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if \"20×\" applies to a different dimension than tokens.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "claude-opus-4-6": {
    "id": "claude-opus-4-6",
    "name": "Claude Opus 4.6 (Non-reasoning, High Effort)",
    "provider": "Anthropic",
    "releaseDate": "2026-02-05",
    "modelCard": "https://artificialanalysis.ai/models/claude-opus-4-6",
    "benchmarks": {
      "aa_intelligence": {
        "score": 46.5,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "gpqa_diamond": {
        "score": 84,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "arena_text": {
        "score": 1498,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1544,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 47.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "hle": {
        "score": 18.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "scicode": {
        "score": 45.7,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "aa_lcr": {
        "score": 58.3,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "tau2_bench": {
        "score": 84.8,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "terminalbench_hard": {
        "score": 48.5,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "ifbench": {
        "score": 44.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "gdpval_aa": {
        "score": 1594.3,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "aa_omniscience": {
        "score": 3.5,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      },
      "critpt": {
        "score": 2.8,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
      }
    },
    "api": {
      "inputPer1M": 6.25,
      "outputPer1M": 25,
      "tokensPerSecAPI": 44,
      "source": "https://artificialanalysis.ai/models/claude-opus-4-6"
    },
    "local": null,
    "subscriptions": {
      "claude_max_20x": {
        "name": "Claude Max 20×",
        "monthlyPrice": 200,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Max 20×. Multi-flip method (3 weekly-% crossings, first discarded). First real Opus-on-Max measurement; prior Opus rows used the Sonnet-derived 203M as a placeholder. Source: claude_measurement_20260424_094106.json.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_094106.json",
        "tokensPerWeek": 248361650,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_094106.json",
        "measurementTaskRuns": null,
        "measurementDelta5h": 18,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)"
      },
      "claude_pro": {
        "name": "Claude Pro",
        "monthlyPrice": 20,
        "confidence": "medium",
        "tokensPerWeek": 15595812,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_161653.json",
        "measurementTaskRuns": 20,
        "measurementDelta5h": 100,
        "measurementDeltaWeekly": 8,
        "measurement_method": "coarse single-delta (20-parallel batch moved weekly 0%→8% in one step; multi-flip data invalidated by all 3 crossings occurring in same batch — known script limitation with large PARALLEL on small-quota plans)",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Pro. First real Pro measurement (Opus specifically). Coarse single-delta method: 1.25M tokens consumed 8% of weekly quota, extrapolated to ~15.6M tok/wk. Session meter went 0→100% confirming Pro has much tighter quota than Max. Sonnet-on-Pro measurement pending (weekly will reset Apr 28). Source: claude_measurement_20260424_161653.json.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_161653.json"
      },
      "claude_max_5x": {
        "name": "Claude Max 5×",
        "monthlyPrice": 100,
        "tokensPerWeek": 62090413,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "claude_max_20x",
          "base_tokens_per_week": 248361650,
          "base_measured_on_model": "Opus 4.7",
          "multiplier": 0.25,
          "multiplier_label": "5/20",
          "source": "https://www.anthropic.com/pricing"
        },
        "confidence": "low",
        "notes": "Anthropic's Max plans are tiered as 5× and 20× Pro. We measure Max 20×; Max 5× is derived as 5/20 of the measured Max 20× number for the same model. Not directly measured.",
        "source": "https://www.anthropic.com/pricing"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "claude-sonnet-4-6": {
    "id": "claude-sonnet-4-6",
    "name": "Claude Sonnet 4.6 (Non-reasoning, High Effort)",
    "provider": "Anthropic",
    "releaseDate": "2026-02-17",
    "modelCard": "https://artificialanalysis.ai/models/claude-sonnet-4-6",
    "benchmarks": {
      "aa_intelligence": {
        "score": 44.4,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "gpqa_diamond": {
        "score": 79.9,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "arena_text": {
        "score": 1466,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1524,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 46.4,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "hle": {
        "score": 13.2,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "scicode": {
        "score": 46.9,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "aa_lcr": {
        "score": 57.7,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "tau2_bench": {
        "score": 79.5,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "terminalbench_hard": {
        "score": 46.2,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "ifbench": {
        "score": 41.2,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "gdpval_aa": {
        "score": 1585.4,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "aa_omniscience": {
        "score": -2.9,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      },
      "critpt": {
        "score": 0.9,
        "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
      }
    },
    "api": {
      "inputPer1M": 3.75,
      "outputPer1M": 15,
      "tokensPerSecAPI": 51.6,
      "source": "https://artificialanalysis.ai/models/claude-sonnet-4-6"
    },
    "local": null,
    "subscriptions": {
      "claude_pro": {
        "name": "Claude Pro",
        "monthlyPrice": 20,
        "confidence": "high",
        "notes": "Measured 2026-04-24 21:39 via claude-code --model sonnet on Sonnet 4.6 on Claude Pro. Multi-flip method, 3 flips in batches 4/7/11, per-1% deltas 168K + 224K (mean 196K/1%). 34% of 5h session consumed, 3% of weekly consumed. Clean run with no same-batch flip pileup. First proper Pro + Sonnet measurement. Source: claude_measurement_20260424_213954.json [auto-update 2026-04-30: tokensPerWeek=17,389,800 from claude_measurement_20260425_091649.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/claude_measurement_20260425_091649.json",
        "tokensPerWeek": 17389800,
        "estimated": false,
        "measuredOnModel": "Sonnet 4.6",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_213954.json",
        "measurementTaskRuns": 33,
        "measurementDelta5h": 34,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded)"
      },
      "claude_max_20x": {
        "name": "Claude Max 20×",
        "monthlyPrice": 200,
        "confidence": "high",
        "tokensPerWeek": 387831600,
        "estimated": false,
        "measuredOnModel": "Sonnet 4.6",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_110655.json",
        "measurementTaskRuns": null,
        "measurementDelta5h": 16,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)",
        "notes": "Measured 2026-04-24 via claude-code --model sonnet on Sonnet 4.6 on Claude Max 20×. Multi-flip method (3 weekly-% crossings, first discarded). Source: claude_measurement_20260424_110655.json. Supersedes the Apr 16 Sonnet 4.5 single-flip run which gave 203M. [auto-update 2026-04-30: tokensPerWeek=387,831,600 from claude_measurement_20260424_110655.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/claude_measurement_20260424_110655.json"
      },
      "claude_max_5x": {
        "name": "Claude Max 5×",
        "monthlyPrice": 100,
        "tokensPerWeek": 96957900,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "claude_max_20x",
          "base_tokens_per_week": 387831600,
          "base_measured_on_model": "Sonnet 4.6",
          "multiplier": 0.25,
          "multiplier_label": "5/20",
          "source": "https://www.anthropic.com/pricing"
        },
        "confidence": "low",
        "notes": "Anthropic's Max plans are tiered as 5× and 20× Pro. We measure Max 20×; Max 5× is derived as 5/20 of the measured Max 20× number for the same model. Not directly measured.",
        "source": "https://www.anthropic.com/pricing"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "claude-4-5-haiku": {
    "id": "claude-4-5-haiku",
    "name": "Claude 4.5 Haiku (Non-reasoning)",
    "provider": "Anthropic",
    "releaseDate": "2025-10-15",
    "modelCard": "https://artificialanalysis.ai/models/claude-4-5-haiku",
    "benchmarks": {
      "aa_intelligence": {
        "score": 31.1,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "gpqa_diamond": {
        "score": 64.6,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "livecodebench": {
        "score": 51.1,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "arena_text": {
        "score": 1409,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1318,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 29.6,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "aa_math_index": {
        "score": 39,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "hle": {
        "score": 4.3,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "scicode": {
        "score": 34.4,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "aa_lcr": {
        "score": 43.7,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "tau2_bench": {
        "score": 32.5,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "terminalbench_hard": {
        "score": 27.3,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "ifbench": {
        "score": 42,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "mmlu_pro": {
        "score": 80,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "aime_25": {
        "score": 39,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "gdpval_aa": {
        "score": 1137.9,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "aa_omniscience": {
        "score": -7.7,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      },
      "critpt": {
        "score": 0,
        "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
      }
    },
    "api": {
      "inputPer1M": 1.25,
      "outputPer1M": 5,
      "tokensPerSecAPI": 100.2,
      "source": "https://artificialanalysis.ai/models/claude-4-5-haiku"
    },
    "local": null,
    "subscriptions": {},
    "_aa_synced": "2026-05-11"
  },
  "claude-4-5-sonnet": {
    "id": "claude-4-5-sonnet",
    "name": "Claude 4.5 Sonnet (Non-reasoning)",
    "provider": "Anthropic",
    "releaseDate": "2025-09-29",
    "modelCard": "https://artificialanalysis.ai/models/claude-4-5-sonnet",
    "benchmarks": {
      "aa_intelligence": {
        "score": 37.1,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "gpqa_diamond": {
        "score": 72.7,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "livecodebench": {
        "score": 59,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "arena_text": {
        "score": 1453,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1386,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 33.5,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "aa_math_index": {
        "score": 37,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "hle": {
        "score": 7.1,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "scicode": {
        "score": 42.8,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "aa_lcr": {
        "score": 51.3,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "tau2_bench": {
        "score": 70.5,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "terminalbench_hard": {
        "score": 28.8,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "ifbench": {
        "score": 42.7,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "mmlu_pro": {
        "score": 86,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "aime_25": {
        "score": 37,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "gdpval_aa": {
        "score": 1314.2,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "aa_omniscience": {
        "score": -9.2,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      },
      "critpt": {
        "score": 0,
        "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
      }
    },
    "api": {
      "inputPer1M": 3.75,
      "outputPer1M": 15,
      "tokensPerSecAPI": 45.1,
      "source": "https://artificialanalysis.ai/models/claude-4-5-sonnet"
    },
    "local": null,
    "subscriptions": {
      "claude_pro": {
        "name": "Claude Pro",
        "monthlyPrice": 20,
        "confidence": "high",
        "notes": "Measured 2026-04-24 21:39 via claude-code --model sonnet on Sonnet 4.6 on Claude Pro. Multi-flip method, 3 flips in batches 4/7/11, per-1% deltas 168K + 224K (mean 196K/1%). 34% of 5h session consumed, 3% of weekly consumed. Clean run with no same-batch flip pileup. First proper Pro + Sonnet measurement. Source: claude_measurement_20260424_213954.json",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_213954.json",
        "tokensPerWeek": 19639400,
        "estimated": false,
        "measuredOnModel": "Sonnet 4.6",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_213954.json",
        "measurementTaskRuns": 33,
        "measurementDelta5h": 34,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded)"
      },
      "claude_max_20x": {
        "name": "Claude Max 20×",
        "monthlyPrice": 200,
        "confidence": "high",
        "tokensPerWeek": 387831650,
        "estimated": false,
        "measuredOnModel": "Sonnet 4.6",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_110655.json",
        "measurementTaskRuns": null,
        "measurementDelta5h": 16,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)",
        "notes": "Measured 2026-04-24 via claude-code --model sonnet on Sonnet 4.6 on Claude Max 20×. Multi-flip method (3 weekly-% crossings, first discarded). Source: claude_measurement_20260424_110655.json. Supersedes the Apr 16 Sonnet 4.5 single-flip run which gave 203M.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/claude_measurement_20260424_110655.json"
      },
      "claude_max_5x": {
        "name": "Claude Max 5×",
        "monthlyPrice": 100,
        "tokensPerWeek": 96957913,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "claude_max_20x",
          "base_tokens_per_week": 387831650,
          "base_measured_on_model": "Sonnet 4.6",
          "multiplier": 0.25,
          "multiplier_label": "5/20",
          "source": "https://www.anthropic.com/pricing"
        },
        "confidence": "low",
        "notes": "Anthropic's Max plans are tiered as 5× and 20× Pro. We measure Max 20×; Max 5× is derived as 5/20 of the measured Max 20× number for the same model. Not directly measured.",
        "source": "https://www.anthropic.com/pricing"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "gpt-5-4": {
    "id": "gpt-5-4",
    "name": "GPT-5.4 (xhigh)",
    "provider": "OpenAI",
    "releaseDate": "2026-03-05",
    "modelCard": "https://artificialanalysis.ai/models/gpt-5-4",
    "benchmarks": {
      "aa_intelligence": {
        "score": 56.8,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "gpqa_diamond": {
        "score": 92,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "arena_text": {
        "score": 1468,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1385,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 57.3,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "hle": {
        "score": 41.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "scicode": {
        "score": 56.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "aa_lcr": {
        "score": 74,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "tau2_bench": {
        "score": 87.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "terminalbench_hard": {
        "score": 57.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "ifbench": {
        "score": 73.9,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "gdpval_aa": {
        "score": 1673.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "aa_omniscience": {
        "score": 5.7,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      },
      "critpt": {
        "score": 23.4,
        "source": "https://artificialanalysis.ai/models/gpt-5-4"
      }
    },
    "api": {
      "inputPer1M": 2.5,
      "outputPer1M": 15,
      "tokensPerSecAPI": 96,
      "source": "https://artificialanalysis.ai/models/gpt-5-4"
    },
    "local": null,
    "subscriptions": {
      "chatgpt_plus": {
        "name": "ChatGPT Plus",
        "monthlyPrice": 20,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via codex-cli on gpt-5.4 xhigh. Multi-flip method (3 weekly-%-left crossings, first discarded). Per-1% tokens: 1,897,399 then 1,906,600 — tight agreement. Source: codex_measurement_20260424_110957.json. Supersedes the Apr 13 single-flip run which gave 13M. [auto-update 2026-04-30: tokensPerWeek=190,200,000 from codex_measurement_20260424_110957.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/codex_measurement_20260424_110957.json",
        "tokensPerWeek": 190200000,
        "estimated": false,
        "measuredOnModel": "gpt-5.4 (reasoning xhigh, summaries auto)",
        "measurementFiles": 7,
        "measurementBestFile": "codex_measurement_20260424_110957.json",
        "measurementTaskRuns": 40,
        "measurementDelta5h": 21,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded due to unknown starting position within bucket)"
      },
      "chatgpt_business": {
        "name": "ChatGPT Business",
        "monthlyPrice": 30,
        "confidence": "medium",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/codex_measurement_20260424_230604.json",
        "notes": "Measured 2026-04-24 23:06 via codex-cli on gpt-5.4 xhigh on Business. Clean multi-flip: 3 flips in 3 separate batches (batches 1/3/4). Per-1% usable deltas: 356,526 and 567,666 — mean 462,096/1%. The 59% spread between the two deltas suggests the 3-flip sample is noisier than ideal; a 5-flip measurement would tighten the estimate. Coarse 3%-span gives 36.8M. Supersedes the contaminated 22:28 measurement (27M, multi-flip broken) and the Apr 15 single-flip measurement (60M, likely high-end outlier). Cache hit rate 80%. [auto-update 2026-04-30: tokensPerWeek=46,209,600 from codex_measurement_20260424_230604.json]",
        "tokensPerWeek": 46209600,
        "estimated": false,
        "measuredOnModel": "gpt-5.4 (reasoning xhigh, summaries auto)",
        "measurementFiles": 1,
        "measurementBestFile": "codex_measurement_20260424_230604.json",
        "measurementTaskRuns": 4,
        "measurementDelta5h": 22,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded)"
      },
      "chatgpt_pro_100": {
        "name": "ChatGPT Pro $100",
        "monthlyPrice": 100,
        "tokensPerWeek": 951000000,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 190200000,
          "base_measured_on_model": "gpt-5.4 (reasoning xhigh, summaries auto)",
          "multiplier": 5,
          "multiplier_label": "5×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $100 plan launched Apr 2026. OpenAI states 5× higher weekly limits than Plus (10× Codex usage through May 31, 2026 as a launch promo). We have not measured Pro $100 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if OpenAI's \"5×\" applies to a dimension other than tokens (e.g. rate-limit slots, message count). Help us verify by running scripts/measure-codex-quota.sh on a Pro $100 account.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      },
      "chatgpt_pro_200": {
        "name": "ChatGPT Pro $200",
        "monthlyPrice": 200,
        "tokensPerWeek": 3804000000,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 190200000,
          "base_measured_on_model": "gpt-5.4 (reasoning xhigh, summaries auto)",
          "multiplier": 20,
          "multiplier_label": "20×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $200 plan. OpenAI states 20× higher weekly limits than Plus (with a 25× Codex 5-hour bonus through May 31, 2026). We have not measured Pro $200 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if \"20×\" applies to a different dimension than tokens.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "gpt-5-3-codex": {
    "id": "gpt-5-3-codex",
    "name": "GPT-5.3 Codex (xhigh)",
    "provider": "OpenAI",
    "releaseDate": "2026-02-05",
    "modelCard": "https://artificialanalysis.ai/models/gpt-5-3-codex",
    "benchmarks": {
      "aa_intelligence": {
        "score": 53.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "gpqa_diamond": {
        "score": 91.5,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "arena_text": {
        "score": 1449,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1372,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 53.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "hle": {
        "score": 39.9,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "scicode": {
        "score": 53.2,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "aa_lcr": {
        "score": 74,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "tau2_bench": {
        "score": 86,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "terminalbench_hard": {
        "score": 53,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "ifbench": {
        "score": 75.4,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "gdpval_aa": {
        "score": 1481.2,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "aa_omniscience": {
        "score": 9.9,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      },
      "critpt": {
        "score": 16.9,
        "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
      }
    },
    "api": {
      "inputPer1M": 1.75,
      "outputPer1M": 14,
      "tokensPerSecAPI": 96.4,
      "source": "https://artificialanalysis.ai/models/gpt-5-3-codex"
    },
    "local": null,
    "subscriptions": {},
    "_aa_synced": "2026-05-11"
  },
  "gpt-5-2-codex": {
    "id": "gpt-5-2-codex",
    "name": "GPT-5.2 Codex (xhigh)",
    "provider": "OpenAI",
    "releaseDate": "2025-12-11",
    "modelCard": "https://artificialanalysis.ai/models/gpt-5-2-codex",
    "benchmarks": {
      "aa_intelligence": {
        "score": 49,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "gpqa_diamond": {
        "score": 89.9,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "arena_text": {
        "score": 1420,
        "source": "https://lmarena.ai/leaderboard",
        "_estimated": true
      },
      "arena_code": {
        "score": 1334,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 43,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "hle": {
        "score": 33.5,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "scicode": {
        "score": 54.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "aa_lcr": {
        "score": 75.7,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "tau2_bench": {
        "score": 92.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "terminalbench_hard": {
        "score": 37.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "ifbench": {
        "score": 77.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "gdpval_aa": {
        "score": 1291.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "aa_omniscience": {
        "score": -2.5,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      },
      "critpt": {
        "score": 8.7,
        "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
      }
    },
    "api": {
      "inputPer1M": 1.75,
      "outputPer1M": 14,
      "tokensPerSecAPI": 95.8,
      "source": "https://artificialanalysis.ai/models/gpt-5-2-codex"
    },
    "local": null,
    "subscriptions": {
      "chatgpt_plus": {
        "name": "ChatGPT Plus",
        "monthlyPrice": 20,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via codex-cli on gpt-5.4 xhigh on Plus. Multi-flip method. Note: this row shows ChatGPT Plus capacity on GPT-5.2 Codex (xhigh), but actual measurement was on GPT-5.4 — older Plus-tier models share the same weekly quota meter so the tokens/week figure applies, but quality is different.",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/codex_measurement_20260424_110957.json",
        "tokensPerWeek": 190199950,
        "estimated": false,
        "measuredOnModel": "gpt-5.4 (reasoning xhigh, summaries auto)",
        "measurementFiles": 7,
        "measurementBestFile": "codex_measurement_20260424_110957.json",
        "measurementTaskRuns": 40,
        "measurementDelta5h": 21,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)"
      },
      "chatgpt_business": {
        "name": "ChatGPT Business",
        "monthlyPrice": 30,
        "confidence": "medium",
        "source": "https://github.com/desktop-commander/best-value-ai/blob/master/measurements/codex_measurement_20260424_230604.json",
        "notes": "Business quota measured on GPT-5.4 (not directly on GPT-5.2 Codex (xhigh)), but weekly quota is per-account shared across models. Measured 2026-04-24 23:06 via codex-cli on gpt-5.4 xhigh on Business. Clean multi-flip: 3 flips in 3 separate batches (batches 1/3/4). Per-1% usable deltas: 356,526 and 567,666 — mean 462,096/1%. The 59% spread between the two deltas suggests the 3-flip sample is noisier than ideal; a 5-flip measurement would tighten the estimate. Coarse 3%-span gives 36.8M. Supersedes the contaminated 22:28 measurement (27M, multi-flip broken) and the Apr 15 single-flip measurement (60M, likely high-end outlier). Cache hit rate 80%.",
        "tokensPerWeek": 46209600,
        "estimated": false,
        "measuredOnModel": "gpt-5.4 (reasoning xhigh, summaries auto)",
        "measurementFiles": 1,
        "measurementBestFile": "codex_measurement_20260424_230604.json",
        "measurementTaskRuns": 4,
        "measurementDelta5h": 22,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded)"
      },
      "chatgpt_pro_100": {
        "name": "ChatGPT Pro $100",
        "monthlyPrice": 100,
        "tokensPerWeek": 950999750,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 190199950,
          "base_measured_on_model": "gpt-5.4 (reasoning xhigh, summaries auto)",
          "multiplier": 5,
          "multiplier_label": "5×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $100 plan launched Apr 2026. OpenAI states 5× higher weekly limits than Plus (10× Codex usage through May 31, 2026 as a launch promo). We have not measured Pro $100 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if OpenAI's \"5×\" applies to a dimension other than tokens (e.g. rate-limit slots, message count). Help us verify by running scripts/measure-codex-quota.sh on a Pro $100 account.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      },
      "chatgpt_pro_200": {
        "name": "ChatGPT Pro $200",
        "monthlyPrice": 200,
        "tokensPerWeek": 3803999000,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 190199950,
          "base_measured_on_model": "gpt-5.4 (reasoning xhigh, summaries auto)",
          "multiplier": 20,
          "multiplier_label": "20×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $200 plan. OpenAI states 20× higher weekly limits than Plus (with a 25× Codex 5-hour bonus through May 31, 2026). We have not measured Pro $200 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if \"20×\" applies to a different dimension than tokens.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "gemini-3-1-pro": {
    "id": "gemini-3-1-pro",
    "name": "Gemini 3.1 Pro Preview",
    "provider": "Google",
    "releaseDate": "2026-02-19",
    "modelCard": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview",
    "benchmarks": {
      "aa_intelligence": {
        "score": 57.2,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "gpqa_diamond": {
        "score": 94.1,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "arena_text": {
        "score": 1492,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1452,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 55.5,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "hle": {
        "score": 44.7,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "scicode": {
        "score": 58.9,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "aa_lcr": {
        "score": 72.7,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "tau2_bench": {
        "score": 95.6,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "terminalbench_hard": {
        "score": 53.8,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "ifbench": {
        "score": 77.1,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "gdpval_aa": {
        "score": 1314.1,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "aa_omniscience": {
        "score": 32.9,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      },
      "critpt": {
        "score": 17.7,
        "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
      }
    },
    "api": {
      "inputPer1M": 2,
      "outputPer1M": 12,
      "tokensPerSecAPI": 130.7,
      "source": "https://artificialanalysis.ai/models/gemini-3-1-pro-preview"
    },
    "local": null,
    "subscriptions": {
      "google_ai_ultra": {
        "name": "Google AI Ultra",
        "monthlyPrice": 250,
        "confidence": "low",
        "notes": "Google AI Ultra (/mo). Estimated ~1000 msgs/day avg turn 6,400 tokens.",
        "source": "https://one.google.com/about/plans",
        "tokensPerWeek": 44800000,
        "estimated": true
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "glm-5": {
    "id": "glm-5",
    "name": "GLM-5 (Reasoning)",
    "provider": "Z AI",
    "releaseDate": "2026-02-11",
    "modelCard": "https://huggingface.co/THUDM/GLM-5",
    "benchmarks": {
      "aa_intelligence": {
        "score": 49.8,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "gpqa_diamond": {
        "score": 82,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "arena_text": {
        "score": 1457,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1436,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 44.2,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "hle": {
        "score": 27.2,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "scicode": {
        "score": 46.2,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "aa_lcr": {
        "score": 63.3,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "tau2_bench": {
        "score": 98.2,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "terminalbench_hard": {
        "score": 43.2,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "ifbench": {
        "score": 72.3,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "gdpval_aa": {
        "score": 1400,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "aa_omniscience": {
        "score": 2,
        "source": "https://artificialanalysis.ai/models/glm-5"
      },
      "critpt": {
        "score": 2,
        "source": "https://artificialanalysis.ai/models/glm-5"
      }
    },
    "api": {
      "inputPer1M": 1,
      "outputPer1M": 3.2,
      "tokensPerSecAPI": 69.2,
      "source": "https://artificialanalysis.ai/models/glm-5"
    },
    "local": {
      "mac_m4_max_128gb": {
        "tokensPerSec": 8,
        "quantization": "Q4_K_M",
        "vramRequired": 128,
        "notes": "744B MoE, 40B active params. Requires 128GB+ unified memory. MIT license.",
        "source": "https://huggingface.co/THUDM/GLM-5"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "glm-4.7": {
    "id": "glm-4.7",
    "name": "GLM-4.7 (Reasoning)",
    "provider": "Z AI",
    "releaseDate": "2025-12-22",
    "modelCard": "https://artificialanalysis.ai/models/glm-4-7",
    "benchmarks": {
      "aa_intelligence": {
        "score": 42.1,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "gpqa_diamond": {
        "score": 85.9,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "livecodebench": {
        "score": 89.4,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "arena_text": {
        "score": 1443,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1440,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 36.3,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "aa_math_index": {
        "score": 95,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "hle": {
        "score": 25.1,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "scicode": {
        "score": 45.1,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "aa_lcr": {
        "score": 64,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "tau2_bench": {
        "score": 95.9,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "terminalbench_hard": {
        "score": 31.8,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "ifbench": {
        "score": 67.9,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "mmlu_pro": {
        "score": 85.6,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "aime_25": {
        "score": 95,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "gdpval_aa": {
        "score": 1189.6,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "aa_omniscience": {
        "score": -34.6,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      },
      "critpt": {
        "score": 1.7,
        "source": "https://artificialanalysis.ai/models/glm-4-7"
      }
    },
    "api": {
      "inputPer1M": 0.6,
      "outputPer1M": 2.2,
      "tokensPerSecAPI": 107.2,
      "source": "https://artificialanalysis.ai/models/glm-4-7"
    },
    "local": {
      "rtx_4090": {
        "tokensPerSec": 45,
        "quantization": "Q4_K_M",
        "vramRequired": 20,
        "notes": "Estimated from GLM-4.7 architecture (similar to 30-40B dense models).",
        "source": "https://github.com/ggml-org/llama.cpp/discussions",
        "_estimated": true
      },
      "mac_m3_max_64gb": {
        "tokensPerSec": 30,
        "quantization": "Q4_K_M",
        "vramRequired": 20,
        "source": "https://github.com/ggml-org/llama.cpp/discussions/4167",
        "_estimated": true
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "minimax-m2.5": {
    "id": "minimax-m2.5",
    "name": "MiniMax-M2.5",
    "provider": "MiniMax",
    "releaseDate": "2026-02-12",
    "modelCard": "https://artificialanalysis.ai/models/minimax-m2-5",
    "benchmarks": {
      "aa_intelligence": {
        "score": 41.9,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "gpqa_diamond": {
        "score": 84.8,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "arena_text": {
        "score": 1395,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1382,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 37.4,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "hle": {
        "score": 19.1,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "scicode": {
        "score": 42.6,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "aa_lcr": {
        "score": 66,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "tau2_bench": {
        "score": 95.3,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "terminalbench_hard": {
        "score": 34.8,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "ifbench": {
        "score": 71.6,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "gdpval_aa": {
        "score": 1185.9,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "aa_omniscience": {
        "score": -39.7,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      },
      "critpt": {
        "score": 1.1,
        "source": "https://artificialanalysis.ai/models/minimax-m2-5"
      }
    },
    "api": {
      "inputPer1M": 0.3,
      "outputPer1M": 1.2,
      "tokensPerSecAPI": 91.1,
      "source": "https://artificialanalysis.ai/models/minimax-m2-5"
    },
    "local": {
      "framework_desktop_128gb": {
        "tokensPerSec": 18,
        "quantization": "TQ1_0",
        "vramRequired": 128,
        "notes": "MoE — aggressive TQ1_0 quant needed to fit in 128GB. CPU BLAS benchmarks on OpenBenchmarking.",
        "source": "https://openbenchmarking.org/test/pts/llama-cpp"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "minimax-m2.7": {
    "id": "minimax-m2.7",
    "name": "MiniMax-M2.7",
    "provider": "MiniMax",
    "releaseDate": "2026-03-18",
    "modelCard": "https://artificialanalysis.ai/models/minimax-m2-7",
    "benchmarks": {
      "aa_intelligence": {
        "score": 49.6,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "gpqa_diamond": {
        "score": 87.4,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "arena_text": {
        "score": 1407,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1407,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 41.9,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "hle": {
        "score": 28.1,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "scicode": {
        "score": 47,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "aa_lcr": {
        "score": 68.7,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "tau2_bench": {
        "score": 84.8,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "terminalbench_hard": {
        "score": 39.4,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "ifbench": {
        "score": 75.7,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "gdpval_aa": {
        "score": 1506.2,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "aa_omniscience": {
        "score": 0.7,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      },
      "critpt": {
        "score": 0.6,
        "source": "https://artificialanalysis.ai/models/minimax-m2-7"
      }
    },
    "api": {
      "inputPer1M": 0.3,
      "outputPer1M": 1.2,
      "tokensPerSecAPI": 49.6,
      "source": "https://artificialanalysis.ai/models/minimax-m2-7"
    },
    "local": {
      "framework_desktop_128gb": {
        "tokensPerSec": 12,
        "quantization": "Q4_K_M",
        "vramRequired": 128,
        "notes": "MoE — requires 128GB+ unified memory.",
        "source": "https://github.com/ggml-org/llama.cpp/discussions/16578",
        "_estimated": true
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "kimi-k2.5": {
    "id": "kimi-k2.5",
    "name": "Kimi K2.5 (Reasoning)",
    "provider": "Kimi",
    "releaseDate": "2026-01-27",
    "modelCard": "https://artificialanalysis.ai/models/kimi-k2-5",
    "benchmarks": {
      "aa_intelligence": {
        "score": 46.8,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "gpqa_diamond": {
        "score": 87.9,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "arena_text": {
        "score": 1449,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1430,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 39.5,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "hle": {
        "score": 29.4,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "scicode": {
        "score": 49,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "aa_lcr": {
        "score": 65.3,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "tau2_bench": {
        "score": 95.9,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "terminalbench_hard": {
        "score": 34.8,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "ifbench": {
        "score": 70.2,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "gdpval_aa": {
        "score": 1284.6,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "aa_omniscience": {
        "score": -8.1,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      },
      "critpt": {
        "score": 3.1,
        "source": "https://artificialanalysis.ai/models/kimi-k2-5"
      }
    },
    "api": {
      "inputPer1M": 0.54,
      "outputPer1M": 2.925,
      "tokensPerSecAPI": 45.5,
      "source": "https://artificialanalysis.ai/models/kimi-k2-5"
    },
    "local": {
      "framework_desktop_128gb": {
        "tokensPerSec": 10,
        "quantization": "Q4_K_M",
        "vramRequired": 128,
        "notes": "MoE 1T/32B active. Extremely large — needs 128GB+ unified memory with heavy quantization.",
        "source": "https://github.com/ggml-org/llama.cpp/discussions/16578",
        "_estimated": true
      },
      "mac_m4_max_128gb": {
        "tokensPerSec": 8,
        "quantization": "Q4_K_M",
        "vramRequired": 128,
        "notes": "MoE 1T/32B active. Estimated — community reports similar MoE models at 8-12 t/s on M4 Max 128GB.",
        "source": "https://github.com/ggml-org/llama.cpp/discussions/4167",
        "_estimated": true
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "deepseek-v3.2": {
    "id": "deepseek-v3.2",
    "name": "DeepSeek V3.2 (Non-reasoning)",
    "provider": "DeepSeek",
    "releaseDate": "2025-12-01",
    "modelCard": "https://artificialanalysis.ai/models/deepseek-v3-2",
    "benchmarks": {
      "aa_intelligence": {
        "score": 32.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "gpqa_diamond": {
        "score": 75.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "livecodebench": {
        "score": 59.3,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "arena_text": {
        "score": 1424,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1332,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 34.6,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "aa_math_index": {
        "score": 59,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "hle": {
        "score": 10.5,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "scicode": {
        "score": 38.7,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "aa_lcr": {
        "score": 39,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "tau2_bench": {
        "score": 78.9,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "terminalbench_hard": {
        "score": 32.6,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "ifbench": {
        "score": 49,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "mmlu_pro": {
        "score": 83.7,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "aime_25": {
        "score": 59,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "gdpval_aa": {
        "score": 882.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "aa_omniscience": {
        "score": -46.7,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      },
      "critpt": {
        "score": 0.9,
        "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
      }
    },
    "api": {
      "inputPer1M": 0.5,
      "outputPer1M": 1.6,
      "tokensPerSecAPI": null,
      "source": "https://artificialanalysis.ai/models/deepseek-v3-2"
    },
    "local": {
      "rtx_4090": {
        "tokensPerSec": 15,
        "quantization": "Q4_K_M",
        "vramRequired": 80,
        "notes": "Requires 2× RTX 4090 (48GB total) or heavy CPU offloading. Single 4090 ~4-6 t/s with offload.",
        "source": "https://github.com/ggml-org/llama.cpp/discussions",
        "_estimated": true
      },
      "framework_desktop_128gb": {
        "tokensPerSec": 14,
        "quantization": "Q4_K_M",
        "vramRequired": 110,
        "notes": "Full unified memory run. Estimated from V3.1 on same hardware class.",
        "source": "https://github.com/ggml-org/llama.cpp/discussions",
        "_estimated": true
      },
      "mac_m4_max_128gb": {
        "tokensPerSec": 12,
        "quantization": "Q4_K_M",
        "vramRequired": 110,
        "notes": "Estimated — 685B MoE fits in 128GB with aggressive quant. ~37B active params.",
        "source": "https://github.com/ggml-org/llama.cpp/discussions/4167",
        "_estimated": true
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "qwen3.5-35b-a3b": {
    "id": "qwen3.5-35b-a3b",
    "name": "Qwen3.5 35B A3B (Reasoning)",
    "provider": "Alibaba",
    "releaseDate": "2026-02-24",
    "modelCard": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b",
    "benchmarks": {
      "aa_intelligence": {
        "score": 37.1,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "gpqa_diamond": {
        "score": 84.5,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "arena_code": {
        "score": 1249,
        "source": "https://arena.ai/leaderboard/code"
      },
      "arena_text": {
        "score": 1397,
        "source": "https://arena.ai/leaderboard/text"
      },
      "aa_coding_index": {
        "score": 30.3,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "hle": {
        "score": 19.7,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "scicode": {
        "score": 37.7,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "aa_lcr": {
        "score": 62.7,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "tau2_bench": {
        "score": 89.2,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "terminalbench_hard": {
        "score": 26.5,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "ifbench": {
        "score": 72.5,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "gdpval_aa": {
        "score": 910.3,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "aa_omniscience": {
        "score": -46.4,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      },
      "critpt": {
        "score": 0.9,
        "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
      }
    },
    "api": {
      "inputPer1M": 0.25,
      "outputPer1M": 2,
      "tokensPerSecAPI": 115.9,
      "source": "https://artificialanalysis.ai/models/qwen3-5-35b-a3b"
    },
    "local": {
      "rtx_4090": {
        "tokensPerSec": 122,
        "quantization": "Q4_K_M",
        "vramRequired": 22,
        "notes": "MoE — only 3B active params per token. Fits in single 24GB GPU.",
        "source": "https://agentnativedev.medium.com/qwen-3-5-35b-a3b-why-your-800-gpu-just-became-a-frontier-class-ai-workstation-63cc4d4ebac1"
      },
      "rtx_3090": {
        "tokensPerSec": 110,
        "quantization": "Q4_K_M",
        "vramRequired": 22,
        "notes": "MoE — only 3B active params per token.",
        "source": "https://agentnativedev.medium.com/qwen-3-5-35b-a3b-why-your-800-gpu-just-became-a-frontier-class-ai-workstation-63cc4d4ebac1"
      },
      "mac_m3_max_64gb": {
        "tokensPerSec": 65,
        "quantization": "Q4_K_M",
        "vramRequired": 22,
        "notes": "Estimated from M3 Ultra 8-bit results scaled for M3 Max bandwidth.",
        "source": "https://www.latent.space/p/ainews-the-high-return-activity-of",
        "_estimated": true
      },
      "mac_m3_max_128gb": {
        "tokensPerSec": 4.4,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 4,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 4 msgs. DB variant: qwen3.5:35b-a3b-coding-nvfp4"
      },
      "mac_m1_max_64gb": {
        "tokensPerSec": 14.2,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 5,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 5 msgs. DB variant: qwen/qwen3.5-35b-a3b"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "qwen3.5-27b": {
    "id": "qwen3.5-27b",
    "name": "Qwen3.5 27B (Reasoning)",
    "provider": "Alibaba",
    "releaseDate": "2026-02-24",
    "modelCard": "https://artificialanalysis.ai/models/qwen3-5-27b",
    "benchmarks": {
      "aa_intelligence": {
        "score": 42.1,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "gpqa_diamond": {
        "score": 85.8,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "arena_code": {
        "score": 1352,
        "source": "https://arena.ai/leaderboard/code"
      },
      "arena_text": {
        "score": 1406,
        "source": "https://arena.ai/leaderboard/text"
      },
      "aa_coding_index": {
        "score": 34.9,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "hle": {
        "score": 22.2,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "scicode": {
        "score": 39.5,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "aa_lcr": {
        "score": 67.3,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "tau2_bench": {
        "score": 93.9,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "terminalbench_hard": {
        "score": 32.6,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "ifbench": {
        "score": 75.6,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "gdpval_aa": {
        "score": 1162.9,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "aa_omniscience": {
        "score": -42,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      },
      "critpt": {
        "score": 0.9,
        "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
      }
    },
    "api": {
      "inputPer1M": 0.3,
      "outputPer1M": 2.4,
      "tokensPerSecAPI": 92.7,
      "source": "https://artificialanalysis.ai/models/qwen3-5-27b"
    },
    "local": {
      "rtx_4090": {
        "tokensPerSec": 58,
        "quantization": "Q4_K_M",
        "vramRequired": 16,
        "notes": "Dense 27B. Community reports 55–65 t/s at Q4_K_M on RTX 4090.",
        "source": "https://localllm.in/blog/llamacpp-vram-requirements-for-local-llms"
      },
      "rtx_3090": {
        "tokensPerSec": 48,
        "quantization": "Q4_K_M",
        "vramRequired": 16,
        "notes": "Dense 27B. Scaled from 4090 for 3090 bandwidth.",
        "source": "https://localllm.in/blog/llamacpp-vram-requirements-for-local-llms",
        "_estimated": true
      },
      "mac_m3_max_64gb": {
        "tokensPerSec": 22,
        "quantization": "Q4_K_M",
        "vramRequired": 16,
        "notes": "MLX 8-bit benchmark on M3 Ultra 512GB gives 21.3 t/s; M3 Max 64GB similar.",
        "source": "https://www.latent.space/p/ainews-the-high-return-activity-of"
      },
      "rx_7900_xtx": {
        "tokensPerSec": 2.9,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 6,
        "_estimated": false,
        "notes": "DC telemetry data: 6 messages, Ryzen 9 3900X, Windows"
      },
      "rtx_4060": {
        "tokensPerSec": 6.5,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 6,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 6 msgs. DB variant: qwen3.5:latest"
      },
      "rtx_3060": {
        "tokensPerSec": 0.5,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 8,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 8 msgs. aggregated from 2 DB model variants: jaahas/qwen3.5-uncensored:latest, qwen3.5:27b-20k"
      },
      "mac_m1_max_64gb": {
        "tokensPerSec": 1.2,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 5,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 5 msgs. DB variant: qwen3.5-27b-ultra-uncensored-heretic-v1"
      },
      "rtx_4060_laptop": {
        "tokensPerSec": 0.1,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 1,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 1 msgs. DB variant: Qwen:latest"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "qwen3.5-122b-a10b": {
    "id": "qwen3.5-122b-a10b",
    "name": "Qwen3.5 122B A10B (Reasoning)",
    "provider": "Alibaba",
    "releaseDate": "2026-02-24",
    "modelCard": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b",
    "benchmarks": {
      "aa_intelligence": {
        "score": 41.6,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "gpqa_diamond": {
        "score": 85.7,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "arena_code": {
        "score": 1363,
        "source": "https://arena.ai/leaderboard/code"
      },
      "arena_text": {
        "score": 1418,
        "source": "https://arena.ai/leaderboard/text"
      },
      "aa_coding_index": {
        "score": 34.7,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "hle": {
        "score": 23.4,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "scicode": {
        "score": 42,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "aa_lcr": {
        "score": 66.7,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "tau2_bench": {
        "score": 93.6,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "terminalbench_hard": {
        "score": 31.1,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "ifbench": {
        "score": 75.7,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "gdpval_aa": {
        "score": 1113,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "aa_omniscience": {
        "score": -39.6,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      },
      "critpt": {
        "score": 0.6,
        "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
      }
    },
    "api": {
      "inputPer1M": 0.4,
      "outputPer1M": 3.2,
      "tokensPerSecAPI": 161.6,
      "source": "https://artificialanalysis.ai/models/qwen3-5-122b-a10b"
    },
    "local": {
      "mac_m4_max_128gb": {
        "tokensPerSec": 55,
        "quantization": "Q4_K_M",
        "vramRequired": 75,
        "notes": "MoE — 10B active params. M5 Max 128GB gives 60.6 t/s; M4 Max estimated slightly lower.",
        "source": "https://www.latent.space/p/ainews-the-high-return-activity-of",
        "_estimated": true
      },
      "mac_m3_max_128gb": {
        "tokensPerSec": 42,
        "quantization": "Q4_K_M",
        "vramRequired": 75,
        "notes": "MoE — 10B active. Based on M3 Ultra 8-bit 42.5 t/s benchmark.",
        "source": "https://www.latent.space/p/ainews-the-high-return-activity-of"
      },
      "framework_desktop_128gb": {
        "tokensPerSec": 38,
        "quantization": "Q4_K_M",
        "vramRequired": 75,
        "notes": "MoE. DGX Spark (similar unified memory architecture) reports ~38 t/s at Q4.",
        "source": "https://github.com/ggml-org/llama.cpp/discussions/16578",
        "_estimated": true
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "gemma-4-31b": {
    "name": "Gemma 4 31B",
    "provider": "Google",
    "releaseDate": "2026-02-01",
    "modelCard": "https://huggingface.co/google/gemma-4-31b-it",
    "benchmarks": {
      "aa_intelligence": {
        "score": 39.2,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "arena_text": {
        "score": 1451,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1383,
        "source": "https://arena.ai/leaderboard/code"
      },
      "aa_coding_index": {
        "score": 38.7,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "gpqa_diamond": {
        "score": 85.7,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "hle": {
        "score": 22.7,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "scicode": {
        "score": 43.4,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "aa_lcr": {
        "score": 62,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "tau2_bench": {
        "score": 59.9,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "terminalbench_hard": {
        "score": 36.4,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "ifbench": {
        "score": 75.6,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "gdpval_aa": {
        "score": 1116.3,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "aa_omniscience": {
        "score": -45.4,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      },
      "critpt": {
        "score": 1.4,
        "source": "https://artificialanalysis.ai/models/gemma-4-31b"
      }
    },
    "api": {
      "inputPer1M": 0.14,
      "outputPer1M": 0.4,
      "source": "https://openrouter.ai/google/gemma-4-31b-it"
    },
    "local": {
      "rtx_4090": {
        "tokensPerSec": 19.5,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 1,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 1 msgs. DB variant: gemma4:latest"
      },
      "mac_m3_max_64gb": {
        "tokensPerSec": 28,
        "quantization": "Q4_K_M",
        "vramRequired": 20,
        "source": "https://huggingface.co/google/gemma-4-31b-it"
      },
      "rtx_3090": {
        "tokensPerSec": 28,
        "quantization": "Q4_K_M",
        "vramRequired": 20,
        "source": "https://till-freitag.com/en/blog/open-source-llm-comparison",
        "_estimated": true
      },
      "mac_m4_48gb": {
        "tokensPerSec": 32,
        "quantization": "Q4_K_M",
        "vramRequired": 20,
        "source": "https://github.com/ggml-org/llama.cpp/discussions/4167",
        "_estimated": true
      },
      "macbook_pro_2019_i9": {
        "tokensPerSec": 21.9,
        "quantization": "cloud (mixed)",
        "vramRequired": 64,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 2,
        "_estimated": false,
        "notes": "DC telemetry data: 2 messages, i9-9980HK, macOS"
      },
      "rtx_500_ada_laptop": {
        "tokensPerSec": 14.6,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 1,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 1 msgs. DB variant: gemma4:latest"
      },
      "mac_m1_pro": {
        "tokensPerSec": 6.2,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 8,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 8 msgs. aggregated from 2 DB model variants: gemma4:latest, gemma4:latest"
      },
      "mac_m2_8gb": {
        "tokensPerSec": 2.1,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 7,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 7 msgs. DB variant: gemma4:latest"
      },
      "rtx_2000_ada": {
        "tokensPerSec": 2.1,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 3,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 3 msgs. DB variant: gemma4:latest"
      },
      "mac_m2_pro": {
        "tokensPerSec": 0.7,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 2,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 2 msgs. DB variant: gemma4:latest"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "gemma-4-26b-a4b": {
    "name": "Gemma 4 26B A4B",
    "provider": "Google",
    "releaseDate": "2026-02-01",
    "modelCard": "https://huggingface.co/google/gemma-4-26b-a4b-it",
    "benchmarks": {
      "aa_intelligence": {
        "score": 26.1,
        "source": "https://artificialanalysis.ai/models/gemma-4-26b-a4b-it"
      },
      "arena_text": {
        "score": 1438,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1360,
        "source": "https://arena.ai/leaderboard/code"
      },
      "gdpval_aa": {
        "score": 1013.5,
        "source": "https://artificialanalysis.ai/models/gemma-4-26b-a4b"
      },
      "aa_omniscience": {
        "score": -48.1,
        "source": "https://artificialanalysis.ai/models/gemma-4-26b-a4b"
      },
      "critpt": {
        "score": 0,
        "source": "https://artificialanalysis.ai/models/gemma-4-26b-a4b"
      }
    },
    "api": null,
    "local": {
      "rtx_4090": {
        "tokensPerSec": 1.2,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 1,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 1 msgs. DB variant: gemma-4-26b-a4b-it"
      },
      "mac_m3_max_64gb": {
        "tokensPerSec": 45,
        "quantization": "Q4_K_M",
        "vramRequired": 8,
        "source": "https://huggingface.co/google/gemma-4-26b-a4b-it"
      },
      "rtx_5090": {
        "tokensPerSec": 85,
        "quantization": "Q4_K_M",
        "vramRequired": 8,
        "notes": "MoE — only 4B active. Extremely fast for its capability level.",
        "source": "https://till-freitag.com/en/blog/open-source-llm-comparison"
      },
      "rtx_3090": {
        "tokensPerSec": 48,
        "quantization": "Q4_K_M",
        "vramRequired": 8,
        "source": "https://github.com/ggml-org/llama.cpp/discussions",
        "_estimated": true
      },
      "mac_m1_max_64gb": {
        "tokensPerSec": 6.4,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 5,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 5 msgs. DB variant: google/gemma-4-26b-a4b"
      },
      "rtx_3080_ti": {
        "tokensPerSec": 5.6,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 1,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 1 msgs. DB variant: google/gemma-4-26b-a4b"
      },
      "rtx_4060": {
        "tokensPerSec": 8.7,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 5,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 5 msgs. DB variant: bjoernb/gemma4-e4b-think:latest"
      },
      "rtx_5070": {
        "tokensPerSec": 9.4,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 15,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 15 msgs. DB variant: gemma4:e4b"
      },
      "intel_uhd_615_laptop": {
        "tokensPerSec": 10.1,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 4,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 4 msgs. DB variant: gemma4:26b"
      },
      "rtx_5060_ti": {
        "tokensPerSec": 11.5,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 10,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 10 msgs. DB variant: gemma4:e4b"
      },
      "mac_m4_pro": {
        "tokensPerSec": 4.9,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 4,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 4 msgs. DB variant: gemma4:e4b"
      },
      "gtx_1660_super": {
        "tokensPerSec": 1.2,
        "source": "https://github.com/desktop-commander/best-value-ai#dc-telemetry-data",
        "_dcMsgs": 2,
        "_estimated": false,
        "notes": "DC telemetry data (2026-04-17): 2 msgs. DB variant: google/gemma-4-e4b"
      }
    },
    "subscriptions": null,
    "_aa_synced": "2026-04-03"
  },
  "claude-opus-4-7": {
    "id": "claude-opus-4-7",
    "name": "Claude Opus 4.7 (Non-reasoning, High Effort)",
    "provider": "Anthropic",
    "releaseDate": "2026-04-16",
    "modelCard": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning",
    "benchmarks": {
      "aa_intelligence": {
        "score": 51.8,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "aa_coding_index": {
        "score": 53.1,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "gpqa_diamond": {
        "score": 88.5,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "hle": {
        "score": 31.2,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "ifbench": {
        "score": 43.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "aa_lcr": {
        "score": 67,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "scicode": {
        "score": 50.1,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "tau2_bench": {
        "score": 74,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "terminalbench_hard": {
        "score": 54.5,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "arena_text": {
        "score": 1491,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1560,
        "source": "https://arena.ai/leaderboard/code"
      },
      "gdpval_aa": {
        "score": 1689.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "aa_omniscience": {
        "score": 14.2,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      },
      "critpt": {
        "score": 5.1,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
      }
    },
    "api": {
      "inputPer1M": 6.25,
      "outputPer1M": 25,
      "tokensPerSecAPI": 53.3,
      "source": "https://artificialanalysis.ai/models/claude-opus-4-7-non-reasoning"
    },
    "local": null,
    "subscriptions": {
      "claude_max_20x": {
        "name": "Claude Max 20×",
        "monthlyPrice": 200,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Max 20×. Multi-flip method (3 weekly-% crossings, first discarded). First real Opus-on-Max measurement; prior Opus rows used the Sonnet-derived 203M as a placeholder. Source: claude_measurement_20260424_094106.json. [auto-update 2026-04-30: tokensPerWeek=162,139,500 from claude_measurement_20260429_195833.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/claude_measurement_20260429_195833.json",
        "tokensPerWeek": 162139500,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_094106.json",
        "measurementTaskRuns": null,
        "measurementDelta5h": 18,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)"
      },
      "claude_pro": {
        "name": "Claude Pro",
        "monthlyPrice": 20,
        "confidence": "high",
        "tokensPerWeek": 10180000,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_161653.json",
        "measurementTaskRuns": 20,
        "measurementDelta5h": 100,
        "measurementDeltaWeekly": 8,
        "measurement_method": "coarse single-delta (20-parallel batch moved weekly 0%→8% in one step; multi-flip data invalidated by all 3 crossings occurring in same batch — known script limitation with large PARALLEL on small-quota plans)",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Pro. First real Pro measurement (Opus specifically). Coarse single-delta method: 1.25M tokens consumed 8% of weekly quota, extrapolated to ~15.6M tok/wk. Session meter went 0→100% confirming Pro has much tighter quota than Max. Sonnet-on-Pro measurement pending (weekly will reset Apr 28). Source: claude_measurement_20260424_161653.json. [auto-update 2026-04-30: tokensPerWeek=10,180,000 from claude_measurement_20260429_185351.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/claude_measurement_20260429_185351.json"
      },
      "claude_max_5x": {
        "name": "Claude Max 5×",
        "monthlyPrice": 100,
        "tokensPerWeek": 40534875,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "claude_max_20x",
          "base_tokens_per_week": 162139500,
          "base_measured_on_model": "Opus 4.7",
          "multiplier": 0.25,
          "multiplier_label": "5/20",
          "source": "https://www.anthropic.com/pricing"
        },
        "confidence": "low",
        "notes": "Anthropic's Max plans are tiered as 5× and 20× Pro. We measure Max 20×; Max 5× is derived as 5/20 of the measured Max 20× number for the same model. Not directly measured.",
        "source": "https://www.anthropic.com/pricing"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "claude-opus-4-7-thinking": {
    "id": "claude-opus-4-7-thinking",
    "name": "Claude Opus 4.7 (Reasoning, Max Effort)",
    "provider": "Anthropic",
    "releaseDate": "2026-04-16",
    "modelCard": "https://artificialanalysis.ai/models/claude-opus-4-7",
    "benchmarks": {
      "aa_intelligence": {
        "score": 57.3,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "aa_coding_index": {
        "score": 52.5,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "gpqa_diamond": {
        "score": 91.4,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "hle": {
        "score": 39.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "ifbench": {
        "score": 58.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "aa_lcr": {
        "score": 70.3,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "scicode": {
        "score": 54.5,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "tau2_bench": {
        "score": 88.6,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "terminalbench_hard": {
        "score": 51.5,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "arena_text": {
        "score": 1503,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1570,
        "source": "https://arena.ai/leaderboard/code"
      },
      "gdpval_aa": {
        "score": 1752.7,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "aa_omniscience": {
        "score": 26.2,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      },
      "critpt": {
        "score": 12,
        "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
      }
    },
    "api": {
      "inputPer1M": 6.25,
      "outputPer1M": 25,
      "tokensPerSecAPI": 66.2,
      "source": "https://artificialanalysis.ai/models/claude-opus-4-7"
    },
    "local": null,
    "subscriptions": {
      "claude_max_20x": {
        "name": "Claude Max 20×",
        "monthlyPrice": 200,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Max 20×. Multi-flip method (3 weekly-% crossings, first discarded). First real Opus-on-Max measurement; prior Opus rows used the Sonnet-derived 203M as a placeholder. Source: claude_measurement_20260424_094106.json. [auto-update 2026-04-30: tokensPerWeek=162,139,500 from claude_measurement_20260429_195833.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/claude_measurement_20260429_195833.json",
        "tokensPerWeek": 162139500,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_094106.json",
        "measurementTaskRuns": null,
        "measurementDelta5h": 18,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-meter crossings, first flip discarded)"
      },
      "claude_pro": {
        "name": "Claude Pro",
        "monthlyPrice": 20,
        "confidence": "high",
        "tokensPerWeek": 10180000,
        "estimated": false,
        "measuredOnModel": "Opus 4.7",
        "measurementFiles": 1,
        "measurementBestFile": "claude_measurement_20260424_161653.json",
        "measurementTaskRuns": 20,
        "measurementDelta5h": 100,
        "measurementDeltaWeekly": 8,
        "measurement_method": "coarse single-delta (20-parallel batch moved weekly 0%→8% in one step; multi-flip data invalidated by all 3 crossings occurring in same batch — known script limitation with large PARALLEL on small-quota plans)",
        "notes": "Measured 2026-04-24 via claude-code --model opus on Opus 4.7 on Claude Pro. First real Pro measurement (Opus specifically). Coarse single-delta method: 1.25M tokens consumed 8% of weekly quota, extrapolated to ~15.6M tok/wk. Session meter went 0→100% confirming Pro has much tighter quota than Max. Sonnet-on-Pro measurement pending (weekly will reset Apr 28). Source: claude_measurement_20260424_161653.json. [auto-update 2026-04-30: tokensPerWeek=10,180,000 from claude_measurement_20260429_185351.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/claude_measurement_20260429_185351.json"
      },
      "claude_max_5x": {
        "name": "Claude Max 5×",
        "monthlyPrice": 100,
        "tokensPerWeek": 40534875,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "claude_max_20x",
          "base_tokens_per_week": 162139500,
          "base_measured_on_model": "Opus 4.7",
          "multiplier": 0.25,
          "multiplier_label": "5/20",
          "source": "https://www.anthropic.com/pricing"
        },
        "confidence": "low",
        "notes": "Anthropic's Max plans are tiered as 5× and 20× Pro. We measure Max 20×; Max 5× is derived as 5/20 of the measured Max 20× number for the same model. Not directly measured.",
        "source": "https://www.anthropic.com/pricing"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "gpt-5-5": {
    "id": "gpt-5-5",
    "name": "GPT-5.5 (xhigh)",
    "provider": "OpenAI",
    "releaseDate": "2026-04-23",
    "modelCard": "https://artificialanalysis.ai/models/gpt-5-5",
    "benchmarks": {
      "aa_intelligence": {
        "score": 60.2,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "aa_coding_index": {
        "score": 59.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "gpqa_diamond": {
        "score": 93.5,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "hle": {
        "score": 44.3,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "ifbench": {
        "score": 75.9,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "aa_lcr": {
        "score": 74.3,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "scicode": {
        "score": 56.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "tau2_bench": {
        "score": 93.9,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "terminalbench_hard": {
        "score": 60.6,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "gdpval_aa": {
        "score": 1774.2,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "aa_omniscience": {
        "score": 20.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "critpt": {
        "score": 27.1,
        "source": "https://artificialanalysis.ai/models/gpt-5-5"
      },
      "arena_text": {
        "score": 1484,
        "source": "https://arena.ai/leaderboard/text"
      },
      "arena_code": {
        "score": 1491,
        "source": "https://arena.ai/leaderboard/code"
      }
    },
    "api": {
      "inputPer1M": 5,
      "outputPer1M": 30,
      "tokensPerSecAPI": 62.9,
      "source": "https://artificialanalysis.ai/models/gpt-5-5"
    },
    "local": null,
    "subscriptions": {
      "chatgpt_plus": {
        "name": "ChatGPT Plus",
        "monthlyPrice": 20,
        "confidence": "high",
        "notes": "Measured 2026-04-24 via codex-cli on gpt-5.5 xhigh on Plus. Multi-flip method, 2 usable 1% crossings (843K and 1,065K tokens per 1%). File reconstructed from /tmp artifacts after original script exited before writing output. A cleaner 5-flip rerun is recommended to tighten the estimate. [auto-update 2026-04-30: tokensPerWeek=36,955,000 from codex_measurement_20260429_190635.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/codex_measurement_20260429_190635.json",
        "tokensPerWeek": 36955000,
        "estimated": false,
        "measuredOnModel": "gpt-5.5 (reasoning xhigh, summaries auto)",
        "measurementFiles": 1,
        "measurementBestFile": "codex_measurement_20260424_194218.json",
        "measurementTaskRuns": 11,
        "measurementDelta5h": 15,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded; 2 usable deltas)"
      },
      "chatgpt_business": {
        "name": "ChatGPT Business",
        "monthlyPrice": 30,
        "confidence": "high",
        "notes": "Measured 2026-04-24 22:38 via codex-cli on gpt-5.5 xhigh on Business. Clean multi-flip: 3 flips in 3 separate batches, per-1% deltas of 181,969 + 181,421 + 179,883 — <1% spread across flips. Cache hit rate 67%. [auto-update 2026-04-30: tokensPerWeek=18,065,200 from codex_measurement_20260424_223823.json]",
        "source": "https://github.com/wonderwhy-er/best-value-ai/blob/master/measurements/codex_measurement_20260424_223823.json",
        "tokensPerWeek": 18065200,
        "estimated": false,
        "measuredOnModel": "gpt-5.5 (reasoning xhigh, summaries auto)",
        "measurementFiles": 1,
        "measurementBestFile": "codex_measurement_20260424_223823.json",
        "measurementTaskRuns": 3,
        "measurementDelta5h": 21,
        "measurementDeltaWeekly": 3,
        "measurement_method": "multi-flip (3 weekly-% crossings, first discarded)"
      },
      "chatgpt_pro_100": {
        "name": "ChatGPT Pro $100",
        "monthlyPrice": 100,
        "tokensPerWeek": 184775000,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 36955000,
          "base_measured_on_model": "gpt-5.5 (reasoning xhigh, summaries auto)",
          "multiplier": 5,
          "multiplier_label": "5×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $100 plan launched Apr 2026. OpenAI states 5× higher weekly limits than Plus (10× Codex usage through May 31, 2026 as a launch promo). We have not measured Pro $100 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if OpenAI's \"5×\" applies to a dimension other than tokens (e.g. rate-limit slots, message count). Help us verify by running scripts/measure-codex-quota.sh on a Pro $100 account.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      },
      "chatgpt_pro_200": {
        "name": "ChatGPT Pro $200",
        "monthlyPrice": 200,
        "tokensPerWeek": 739100000,
        "estimated": true,
        "derived": true,
        "derived_from": {
          "base_plan": "chatgpt_plus",
          "base_tokens_per_week": 36955000,
          "base_measured_on_model": "gpt-5.5 (reasoning xhigh, summaries auto)",
          "multiplier": 20,
          "multiplier_label": "20×",
          "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
        },
        "confidence": "low",
        "notes": "Pro $200 plan. OpenAI states 20× higher weekly limits than Plus (with a 25× Codex 5-hour bonus through May 31, 2026). We have not measured Pro $200 directly — this row is derived from our Plus measurement × the vendor-stated multiplier. Real number could differ if \"20×\" applies to a different dimension than tokens.",
        "source": "https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers"
      }
    },
    "_aa_synced": "2026-05-11"
  },
  "deepseek-v4-pro-reasoning": {
    "id": "deepseek-v4-pro-reasoning",
    "name": "DeepSeek V4 Pro (Reasoning, Max Effort)",
    "provider": "DeepSeek",
    "releaseDate": "2026-04-24",
    "modelCard": "https://artificialanalysis.ai/models/deepseek-v4-pro",
    "benchmarks": {
      "aa_intelligence": {
        "score": 51.5,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "aa_coding_index": {
        "score": 47.5,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "gpqa_diamond": {
        "score": 88.8,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "hle": {
        "score": 35.9,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "scicode": {
        "score": 50,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "aa_lcr": {
        "score": 66.3,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "tau2_bench": {
        "score": 96.2,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "terminalbench_hard": {
        "score": 46.2,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "ifbench": {
        "score": 76.5,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "gdpval_aa": {
        "score": 1553.8,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "aa_omniscience": {
        "score": -10,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "critpt": {
        "score": 12.9,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
      },
      "arena_text": {
        "score": 1463,
        "source": "https://arena.ai/leaderboard/text"
      }
    },
    "api": {
      "inputPer1M": 1.74,
      "outputPer1M": 3.48,
      "tokensPerSecAPI": 30.7,
      "source": "https://artificialanalysis.ai/models/deepseek-v4-pro"
    },
    "local": null,
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "deepseek-v4-pro": {
    "id": "deepseek-v4-pro",
    "name": "DeepSeek V4 Pro (Non-reasoning)",
    "provider": "DeepSeek",
    "releaseDate": "2026-04-24",
    "modelCard": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning",
    "benchmarks": {
      "aa_intelligence": {
        "score": 39.3,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "aa_coding_index": {
        "score": 38.4,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "gpqa_diamond": {
        "score": 71.7,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "hle": {
        "score": 7.7,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "scicode": {
        "score": 42.4,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "aa_lcr": {
        "score": 45,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "tau2_bench": {
        "score": 91.2,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "terminalbench_hard": {
        "score": 36.4,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "ifbench": {
        "score": 45.8,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "gdpval_aa": {
        "score": 1480.5,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "aa_omniscience": {
        "score": -30.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "critpt": {
        "score": 0.9,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
      },
      "arena_text": {
        "score": 1463,
        "source": "https://arena.ai/leaderboard/text"
      }
    },
    "api": {
      "inputPer1M": 1.74,
      "outputPer1M": 3.48,
      "tokensPerSecAPI": 30.3,
      "source": "https://artificialanalysis.ai/models/deepseek-v4-pro-non-reasoning"
    },
    "local": null,
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "deepseek-v4-flash-reasoning": {
    "id": "deepseek-v4-flash-reasoning",
    "name": "DeepSeek V4 Flash (Reasoning, Max Effort)",
    "provider": "DeepSeek",
    "releaseDate": "2026-04-24",
    "modelCard": "https://artificialanalysis.ai/models/deepseek-v4-flash",
    "benchmarks": {
      "aa_intelligence": {
        "score": 46.5,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "aa_coding_index": {
        "score": 38.7,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "gpqa_diamond": {
        "score": 89.4,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "hle": {
        "score": 32.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "scicode": {
        "score": 44.9,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "aa_lcr": {
        "score": 63,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "tau2_bench": {
        "score": 95,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "terminalbench_hard": {
        "score": 35.6,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "ifbench": {
        "score": 79.2,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "gdpval_aa": {
        "score": 1388.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "aa_omniscience": {
        "score": -22.9,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "critpt": {
        "score": 7.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
      },
      "arena_text": {
        "score": 1433,
        "source": "https://arena.ai/leaderboard/text"
      }
    },
    "api": {
      "inputPer1M": 0.14,
      "outputPer1M": 0.28,
      "tokensPerSecAPI": 66.2,
      "source": "https://artificialanalysis.ai/models/deepseek-v4-flash"
    },
    "local": null,
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  },
  "deepseek-v4-flash": {
    "id": "deepseek-v4-flash",
    "name": "DeepSeek V4 Flash (Non-reasoning)",
    "provider": "DeepSeek",
    "releaseDate": "2026-04-24",
    "modelCard": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning",
    "benchmarks": {
      "aa_intelligence": {
        "score": 36.5,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "aa_coding_index": {
        "score": 35.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "gpqa_diamond": {
        "score": 71.6,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "hle": {
        "score": 7,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "scicode": {
        "score": 37.3,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "aa_lcr": {
        "score": 33.3,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "tau2_bench": {
        "score": 94.4,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "terminalbench_hard": {
        "score": 34.1,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "ifbench": {
        "score": 47.2,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "gdpval_aa": {
        "score": 1395.6,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "aa_omniscience": {
        "score": -44.2,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "critpt": {
        "score": 0.3,
        "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
      },
      "arena_text": {
        "score": 1433,
        "source": "https://arena.ai/leaderboard/text"
      }
    },
    "api": {
      "inputPer1M": 0.14,
      "outputPer1M": 0.28,
      "tokensPerSecAPI": 66.5,
      "source": "https://artificialanalysis.ai/models/deepseek-v4-flash-non-reasoning"
    },
    "local": null,
    "subscriptions": null,
    "_aa_synced": "2026-05-11"
  }
}