{
  "version": 3,
  "benchmarks": [
    {
      "id": "deepswe",
      "title": "DeepSWE",
      "sourceNote": "Illustrative snapshot from the public DeepSWE leaderboard (mini-swe-agent). Scores and costs are not live OpenCodex metering.",
      "updated": "2026-07-16",
      "taskCount": 113,
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Avg cost per task (USD)",
        "yLabel": "Pass@1",
        "xUnit": "USD",
        "yUnit": "%"
      },
      "rows": [
        {
          "id": "deepswe-gpt-5.6-sol-max",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "effort": "max",
          "score": 73,
          "scoreCi": 3,
          "avgCostUsd": 8.39,
          "outTokens": 60000,
          "steps": 61,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-claude-fable-5-max",
          "model": "claude-fable-5",
          "family": "anthropic",
          "effort": "max",
          "score": 70,
          "scoreCi": 4,
          "avgCostUsd": 21.63,
          "outTokens": 119000,
          "steps": 88,
          "tags": [
            "planner",
            "frontier"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-gpt-5.6-terra-max",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "effort": "max",
          "score": 70,
          "scoreCi": 3,
          "avgCostUsd": 4.95,
          "outTokens": 72000,
          "steps": 76,
          "tags": [
            "workhorse",
            "frontier"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-gpt-5.6-luna-max",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "effort": "max",
          "score": 67,
          "scoreCi": 4,
          "avgCostUsd": 3.03,
          "outTokens": 73000,
          "steps": 102,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-gpt-5.5-xhigh",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "xhigh",
          "score": 67,
          "scoreCi": 6,
          "avgCostUsd": 7.23,
          "outTokens": 46000,
          "steps": 82,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-claude-opus-4.8-max",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "effort": "max",
          "score": 59,
          "scoreCi": 2,
          "avgCostUsd": 13.22,
          "outTokens": 135000,
          "steps": 120,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-claude-sonnet-5-max",
          "model": "claude-sonnet-5",
          "family": "anthropic",
          "effort": "max",
          "score": 54,
          "scoreCi": 4,
          "avgCostUsd": 26.4,
          "outTokens": 214000,
          "steps": 268,
          "tags": [
            "planner"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-grok-4.5-high",
          "model": "grok-4.5",
          "family": "xai",
          "effort": "high",
          "score": 54,
          "scoreCi": 2,
          "avgCostUsd": 2.42,
          "outTokens": 36000,
          "steps": 61,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-muse-spark-1.1-xhigh",
          "model": "muse-spark-1.1",
          "family": "other",
          "effort": "xhigh",
          "score": 53,
          "scoreCi": 3,
          "avgCostUsd": 2.36,
          "outTokens": 74000,
          "steps": 96,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-gpt-5.4-xhigh",
          "model": "gpt-5.4",
          "family": "openai",
          "effort": "xhigh",
          "score": 52,
          "scoreCi": 2,
          "avgCostUsd": 5.65,
          "outTokens": 71000,
          "steps": 70,
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-glm-5.2-max",
          "model": "glm-5.2",
          "family": "zhipu",
          "effort": "max",
          "score": 44,
          "scoreCi": 2,
          "avgCostUsd": 3.92,
          "outTokens": 78000,
          "steps": 129,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-gemini-3.5-flash-medium",
          "model": "gemini-3.5-flash",
          "family": "google",
          "effort": "medium",
          "score": 37,
          "scoreCi": 2,
          "avgCostUsd": 7.34,
          "outTokens": 276000,
          "steps": 86,
          "tags": [
            "fast"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-kimi-k2.7-code",
          "model": "kimi-k2.7-code",
          "family": "moonshot",
          "score": 31,
          "scoreCi": 1,
          "avgCostUsd": 2.82,
          "outTokens": 59000,
          "steps": 149,
          "tags": [
            "cheap-subagent",
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-claude-sonnet-4.6-high",
          "model": "claude-sonnet-4.6",
          "family": "anthropic",
          "effort": "high",
          "score": 30,
          "scoreCi": 4,
          "avgCostUsd": 5.52,
          "outTokens": 76000,
          "steps": 134,
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "deepswe-gemini-3.1-pro-high",
          "model": "gemini-3.1-pro",
          "family": "google",
          "effort": "high",
          "score": 12,
          "scoreCi": 2,
          "avgCostUsd": 9.48,
          "outTokens": 196000,
          "steps": 81,
          "tags": [
            "frontier"
          ],
          "costKind": "measured"
        }
      ],
      "provenance": {
        "url": "https://www.deepwiki.com/deepswe",
        "capturedAt": "2026-07-16",
        "license": "Public leaderboard snapshot; verify current rankings on the source before citing."
      },
      "modeKind": "effort"
    },
    {
      "id": "aa-coding-agent",
      "title": "AA Coding Agent",
      "sourceNote": "Artificial Analysis Coding Agent Index snapshot (post GPT-5.6 GA, 2026-07-09). See provenance.url. Not live OpenCodex metering.",
      "updated": "2026-07-09",
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Cost per task (USD)",
        "yLabel": "Coding Agent Index",
        "xUnit": "USD",
        "yUnit": "index"
      },
      "rows": [
        {
          "id": "aa-composer-2.5-std",
          "model": "composer-2.5",
          "family": "cursor",
          "effort": "std",
          "score": 62,
          "avgCostUsd": 0.07,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-composer-2.5-fast",
          "model": "composer-2.5",
          "family": "cursor",
          "effort": "fast",
          "score": 62,
          "avgCostUsd": 0.44,
          "tags": [
            "workhorse",
            "fast"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-claude-opus-4.7",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "effort": "max",
          "score": 66,
          "avgCostUsd": 4.1,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-gpt-5.5",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "xhigh",
          "score": 65,
          "avgCostUsd": 4.82,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-gpt-5.6-luna",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "effort": "max",
          "score": 75,
          "avgCostUsd": 1.55,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-gpt-5.6-terra",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "effort": "max",
          "score": 77,
          "avgCostUsd": 2.4,
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-gpt-5.6-sol",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "effort": "max",
          "score": 80,
          "avgCostUsd": 3.9,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-grok-4.5",
          "model": "grok-4.5",
          "family": "xai",
          "effort": "high",
          "score": 76,
          "avgCostUsd": 2.54,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-claude-sonnet-4.6",
          "model": "claude-sonnet-4.6",
          "family": "anthropic",
          "effort": "high",
          "score": 58,
          "avgCostUsd": 2.8,
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "aa-claude-fable-5",
          "model": "claude-fable-5",
          "family": "anthropic",
          "effort": "high",
          "score": 64,
          "avgCostUsd": 9.5,
          "tags": [
            "planner",
            "frontier"
          ],
          "costKind": "measured"
        }
      ],
      "provenance": {
        "url": "https://artificialanalysis.ai/agents/coding-agents",
        "capturedAt": "2026-07-09",
        "version": "Coding Agent Index (post GPT-5.6 GA)",
        "license": "Figures from Artificial Analysis public pages; not affiliated with AA."
      },
      "modeKind": "effort"
    },
    {
      "id": "aa-intelligence-index",
      "title": "AA Intelligence Index",
      "sourceNote": "Illustrative snapshot inspired by Artificial Analysis Intelligence Index cost-per-task (Answer / Reasoning / Cache write / Cache hit / Input). Totals match published headline figures where noted; segment splits are approximate for charting. Not live OpenCodex metering.",
      "updated": "2026-07-16",
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Cost per Intelligence Index task (USD)",
        "yLabel": "Intelligence Index",
        "xUnit": "USD",
        "yUnit": "index"
      },
      "rows": [
        {
          "id": "ii-claude-fable-5-max",
          "model": "claude-fable-5",
          "family": "anthropic",
          "effort": "max",
          "score": 60,
          "avgCostUsd": 3.25,
          "costParts": {
            "answer": 0.28,
            "reasoning": 1.72,
            "cacheWrite": 0.52,
            "cacheHit": 0.38,
            "input": 0.35
          },
          "tags": [
            "planner",
            "frontier"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-gpt-5.6-sol-max",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "effort": "max",
          "score": 59,
          "avgCostUsd": 1.04,
          "costParts": {
            "answer": 0.12,
            "reasoning": 0.58,
            "cacheWrite": 0.14,
            "cacheHit": 0.1,
            "input": 0.1
          },
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-gpt-5.6-sol-xhigh",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "effort": "xhigh",
          "score": 58,
          "avgCostUsd": 0.88,
          "costParts": {
            "answer": 0.1,
            "reasoning": 0.48,
            "cacheWrite": 0.12,
            "cacheHit": 0.09,
            "input": 0.09
          },
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-claude-opus-4.8-max",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "effort": "max",
          "score": 57,
          "avgCostUsd": 1.78,
          "costParts": {
            "answer": 0.16,
            "reasoning": 0.92,
            "cacheWrite": 0.3,
            "cacheHit": 0.22,
            "input": 0.18
          },
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-gpt-5.5-xhigh",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "xhigh",
          "score": 56,
          "avgCostUsd": 0.99,
          "costParts": {
            "answer": 0.11,
            "reasoning": 0.54,
            "cacheWrite": 0.08,
            "cacheHit": 0.14,
            "input": 0.12
          },
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-gpt-5.6-terra-max",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "effort": "max",
          "score": 55,
          "avgCostUsd": 0.55,
          "costParts": {
            "answer": 0.06,
            "reasoning": 0.3,
            "cacheWrite": 0.07,
            "cacheHit": 0.06,
            "input": 0.06
          },
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-gpt-5.6-luna-max",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "effort": "max",
          "score": 51,
          "avgCostUsd": 0.21,
          "costParts": {
            "answer": 0.025,
            "reasoning": 0.11,
            "cacheWrite": 0.03,
            "cacheHit": 0.025,
            "input": 0.02
          },
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-gemini-3.1-pro",
          "model": "gemini-3.1-pro",
          "family": "google",
          "effort": "preview",
          "score": 46,
          "avgCostUsd": 0.32,
          "costParts": {
            "answer": 0.04,
            "reasoning": 0.14,
            "cacheWrite": 0.02,
            "cacheHit": 0.06,
            "input": 0.06
          },
          "tags": [
            "frontier",
            "fast"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-deepseek-v4-pro-max",
          "model": "deepseek-v4-pro",
          "family": "deepseek",
          "effort": "max",
          "score": 44,
          "avgCostUsd": 0.04,
          "costParts": {
            "answer": 0.004,
            "reasoning": 0.018,
            "cacheWrite": 0.003,
            "cacheHit": 0.008,
            "input": 0.007
          },
          "tags": [
            "cheap-subagent",
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-grok-4.3-high",
          "model": "grok-4.3",
          "family": "xai",
          "effort": "high",
          "score": 48,
          "avgCostUsd": 0.28,
          "costParts": {
            "answer": 0.04,
            "reasoning": 0.14,
            "cacheWrite": 0.02,
            "cacheHit": 0.04,
            "input": 0.04
          },
          "tags": [
            "workhorse",
            "fast"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-glm-5.2-max",
          "model": "glm-5.2",
          "family": "zhipu",
          "effort": "max",
          "score": 42,
          "avgCostUsd": 0.12,
          "costParts": {
            "answer": 0.015,
            "reasoning": 0.055,
            "cacheWrite": 0.015,
            "cacheHit": 0.02,
            "input": 0.015
          },
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "ii-claude-sonnet-4.6-max",
          "model": "claude-sonnet-4.6",
          "family": "anthropic",
          "effort": "max",
          "score": 49,
          "avgCostUsd": 1.15,
          "costParts": {
            "answer": 0.1,
            "reasoning": 0.55,
            "cacheWrite": 0.22,
            "cacheHit": 0.16,
            "input": 0.12
          },
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        }
      ],
      "provenance": {
        "url": "https://artificialanalysis.ai/leaderboards/models",
        "capturedAt": "2026-07-16",
        "license": "Figures from Artificial Analysis public pages; not affiliated with AA."
      },
      "modeKind": "effort"
    },
    {
      "id": "frontiercode",
      "title": "FrontierCode",
      "sourceNote": "Illustrative snapshot from cognition.com/frontiercode (FrontierCode 1.1 Main, Jul 2026). Score = mergeability rubric; cost = mean USD per rollout. Best vs all reasoning modes match Cognition’s leaderboard toggle. Not live OpenCodex metering.",
      "updated": "2026-07-16",
      "taskCount": 100,
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Avg cost per rollout (USD)",
        "yLabel": "Mergeability score",
        "xUnit": "USD",
        "yUnit": "%"
      },
      "rows": [
        {
          "id": "fc-claude-fable-5-xhigh",
          "model": "claude-fable-5",
          "family": "anthropic",
          "score": 53.5,
          "avgCostUsd": 13.09,
          "outTokens": 58600,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "xhigh",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-fable-5-high",
          "model": "claude-fable-5",
          "family": "anthropic",
          "score": 52.7,
          "avgCostUsd": 9.48,
          "outTokens": 43500,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-fable-5-max",
          "model": "claude-fable-5",
          "family": "anthropic",
          "score": 51.6,
          "avgCostUsd": 19.07,
          "outTokens": 80800,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "max",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-fable-5-medium",
          "model": "claude-fable-5",
          "family": "anthropic",
          "score": 49.8,
          "avgCostUsd": 7.15,
          "outTokens": 33200,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-fable-5-low",
          "model": "claude-fable-5",
          "family": "anthropic",
          "score": 48,
          "avgCostUsd": 4.92,
          "outTokens": 23600,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-sol-max",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "score": 47.5,
          "avgCostUsd": 6.3,
          "outTokens": 33200,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "max",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-sol-xhigh",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "score": 46.8,
          "avgCostUsd": 5,
          "outTokens": 26300,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "xhigh",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.8-max",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "score": 46.5,
          "avgCostUsd": 9.62,
          "outTokens": 95900,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "max",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.8-xhigh",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "score": 45.5,
          "avgCostUsd": 6.78,
          "outTokens": 65700,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "xhigh",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-sol-high",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "score": 45.1,
          "avgCostUsd": 4.1,
          "outTokens": 21900,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.5-xhigh",
          "model": "gpt-5.5",
          "family": "openai",
          "score": 43,
          "avgCostUsd": 4.03,
          "outTokens": 25000,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "effort": "xhigh",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-sonnet-5-xhigh",
          "model": "claude-sonnet-5",
          "family": "anthropic",
          "score": 42.7,
          "avgCostUsd": 9.06,
          "outTokens": 87900,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "effort": "xhigh",
          "costKind": "measured"
        },
        {
          "id": "fc-grok-4.5-high",
          "model": "grok-4.5",
          "family": "xai",
          "score": 42.4,
          "avgCostUsd": 1.3,
          "outTokens": 15300,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-sonnet-5-max",
          "model": "claude-sonnet-5",
          "family": "anthropic",
          "score": 42.4,
          "avgCostUsd": 15.4,
          "outTokens": 142600,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "effort": "max",
          "costKind": "measured"
        },
        {
          "id": "fc-swe-1.7",
          "model": "swe-1.7",
          "family": "other",
          "score": 42.3,
          "avgCostUsd": 1.97,
          "outTokens": 64800,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "fc-grok-4.5-medium",
          "model": "grok-4.5",
          "family": "xai",
          "score": 41.9,
          "avgCostUsd": 1.2,
          "outTokens": 14900,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-terra-max",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "score": 41.3,
          "avgCostUsd": 2.3,
          "outTokens": 40500,
          "tags": [
            "workhorse"
          ],
          "effort": "max",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.8-high",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "score": 41,
          "avgCostUsd": 4.28,
          "outTokens": 41300,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.5-high",
          "model": "gpt-5.5",
          "family": "openai",
          "score": 40.6,
          "avgCostUsd": 3.12,
          "outTokens": 18400,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.8-medium",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "score": 40.5,
          "avgCostUsd": 3.7,
          "outTokens": 34900,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-sol-medium",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "score": 39.9,
          "avgCostUsd": 3.1,
          "outTokens": 17000,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-luna-max",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "score": 39.8,
          "avgCostUsd": 1.8,
          "outTokens": 37400,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "effort": "max",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-sonnet-5-high",
          "model": "claude-sonnet-5",
          "family": "anthropic",
          "score": 39.4,
          "avgCostUsd": 5.49,
          "outTokens": 57700,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-luna-xhigh",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "score": 38.9,
          "avgCostUsd": 1.5,
          "outTokens": 31200,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "effort": "xhigh",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-terra-xhigh",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "score": 38.7,
          "avgCostUsd": 1.6,
          "outTokens": 30700,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "effort": "xhigh",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.7-max",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "score": 38.5,
          "avgCostUsd": 9.09,
          "outTokens": 49100,
          "tags": [
            "workhorse"
          ],
          "effort": "max",
          "costKind": "measured"
        },
        {
          "id": "fc-grok-4.5-low",
          "model": "grok-4.5",
          "family": "xai",
          "score": 37.9,
          "avgCostUsd": 0.8,
          "outTokens": 11900,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-terra-high",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "score": 36.9,
          "avgCostUsd": 1.4,
          "outTokens": 25300,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.5-medium",
          "model": "gpt-5.5",
          "family": "openai",
          "score": 36.6,
          "avgCostUsd": 2.35,
          "outTokens": 13700,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-luna-high",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "score": 35.9,
          "avgCostUsd": 1.1,
          "outTokens": 23200,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-sol-low",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "score": 35.4,
          "avgCostUsd": 2.1,
          "outTokens": 12000,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-sonnet-5-medium",
          "model": "claude-sonnet-5",
          "family": "anthropic",
          "score": 35.2,
          "avgCostUsd": 3.43,
          "outTokens": 38300,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.7-high",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "score": 35.2,
          "avgCostUsd": 5.03,
          "outTokens": 30100,
          "tags": [
            "workhorse"
          ],
          "effort": "high",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.8-low",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "score": 35.1,
          "avgCostUsd": 2.68,
          "outTokens": 25200,
          "tags": [
            "frontier",
            "workhorse",
            "planner"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.7-xhigh",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "score": 34.9,
          "avgCostUsd": 6.87,
          "outTokens": 38400,
          "tags": [
            "workhorse"
          ],
          "effort": "xhigh",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-terra-medium",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "score": 33.9,
          "avgCostUsd": 0.9,
          "outTokens": 16300,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.5-low",
          "model": "gpt-5.5",
          "family": "openai",
          "score": 30.6,
          "avgCostUsd": 1.65,
          "outTokens": 8900,
          "tags": [
            "frontier",
            "workhorse",
            "cheap-subagent"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-kimi-k2.7",
          "model": "kimi-k2.7",
          "family": "moonshot",
          "score": 30.1,
          "avgCostUsd": 3.01,
          "outTokens": 43900,
          "tags": [],
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.7-medium",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "score": 28.9,
          "avgCostUsd": 3.22,
          "outTokens": 21900,
          "tags": [
            "workhorse"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-sonnet-5-low",
          "model": "claude-sonnet-5",
          "family": "anthropic",
          "score": 28.7,
          "avgCostUsd": 2.15,
          "outTokens": 24800,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-claude-opus-4.7-low",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "score": 27.6,
          "avgCostUsd": 2.47,
          "outTokens": 17900,
          "tags": [
            "workhorse"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-luna-medium",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "score": 25.7,
          "avgCostUsd": 0.6,
          "outTokens": 12900,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "medium",
          "costKind": "measured"
        },
        {
          "id": "fc-composer-2.5",
          "model": "composer-2.5",
          "family": "cursor",
          "score": 25.6,
          "avgCostUsd": 3.09,
          "outTokens": 16000,
          "tags": [],
          "costKind": "measured"
        },
        {
          "id": "fc-glm-5.2",
          "model": "glm-5.2",
          "family": "zhipu",
          "score": 24.5,
          "avgCostUsd": 2.47,
          "outTokens": 17800,
          "tags": [],
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-terra-low",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "score": 24.1,
          "avgCostUsd": 0.5,
          "outTokens": 9200,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-deepseek-v4-pro",
          "model": "deepseek-v4-pro",
          "family": "deepseek",
          "score": 17.6,
          "avgCostUsd": 1.6,
          "outTokens": 28800,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "fc-gpt-5.6-luna-low",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "score": 15.4,
          "avgCostUsd": 0.2,
          "outTokens": 6800,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "effort": "low",
          "costKind": "measured"
        },
        {
          "id": "fc-minimax-m3",
          "model": "minimax-m3",
          "family": "other",
          "score": 14.7,
          "avgCostUsd": 0.7,
          "outTokens": 34000,
          "tags": [
            "cheap-subagent",
            "fast"
          ],
          "costKind": "measured"
        },
        {
          "id": "fc-inkling-0.99",
          "model": "inkling",
          "family": "other",
          "score": 14,
          "avgCostUsd": 3.6,
          "outTokens": 33600,
          "tags": [],
          "effort": "0.99",
          "costKind": "measured"
        },
        {
          "id": "fc-qwen-3.7-plus",
          "model": "qwen-3.7-plus",
          "family": "other",
          "score": 10.2,
          "avgCostUsd": 0.2,
          "outTokens": 45300,
          "tags": [
            "cheap-subagent",
            "fast"
          ],
          "costKind": "measured"
        }
      ],
      "provenance": {
        "url": "https://cognition.com/frontiercode",
        "capturedAt": "2026-07-16",
        "license": "Public FrontierCode snapshot; verify on cognition.com before citing."
      },
      "modeKind": "effort"
    },
    {
      "id": "frontierswe",
      "title": "FrontierSWE",
      "sourceNote": "Illustrative snapshot from frontierswe.com (Mean@5 dominance). X-axis uses blended public API $/MTok as a relative cost proxy — not measured $/task. Not live OpenCodex metering.",
      "updated": "2026-07-16",
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Relative API cost (blended $/MTok)",
        "yLabel": "Dominance",
        "xUnit": "USD",
        "yUnit": "%"
      },
      "rows": [
        {
          "id": "fswe-claude-fable-5",
          "model": "claude-fable-5",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 89,
          "avgCostUsd": 30,
          "tags": [
            "planner",
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-grok-4.5",
          "model": "grok-4.5",
          "family": "xai",
          "effort": "grok-cli",
          "score": 78,
          "avgCostUsd": 4,
          "tags": [
            "workhorse",
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-claude-opus-4.8",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 73,
          "avgCostUsd": 15,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-glm-5.2",
          "model": "glm-5.2",
          "family": "zhipu",
          "effort": "claude-code",
          "score": 72,
          "avgCostUsd": 1.98,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-gpt-5.5",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "codex",
          "score": 70,
          "avgCostUsd": 17.5,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-claude-opus-4.7",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 61,
          "avgCostUsd": 15,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-claude-opus-4.6",
          "model": "claude-opus-4.6",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 53,
          "avgCostUsd": 15,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-gpt-5.4",
          "model": "gpt-5.4",
          "family": "openai",
          "effort": "codex",
          "score": 51,
          "avgCostUsd": 8.75,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-composer-2.5",
          "model": "composer-2.5",
          "family": "cursor",
          "effort": "cursor-cli",
          "score": 38,
          "avgCostUsd": 1,
          "tags": [
            "workhorse",
            "cheap-subagent",
            "fast"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-gemini-3.1-pro",
          "model": "gemini-3.1-pro",
          "family": "google",
          "effort": "gemini-cli",
          "score": 37,
          "avgCostUsd": 8.75,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-glm-5.1",
          "model": "glm-5.1",
          "family": "zhipu",
          "effort": "claude-code",
          "score": 29,
          "avgCostUsd": 2.9,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-deepseek-v4-pro",
          "model": "deepseek-v4-pro",
          "family": "deepseek",
          "effort": "claude-code",
          "score": 27,
          "avgCostUsd": 2.4,
          "tags": [
            "cheap-subagent",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-kimi-k2.6",
          "model": "kimi-k2.6",
          "family": "moonshot",
          "effort": "kimi-cli",
          "score": 25,
          "avgCostUsd": 2.13,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-kimi-k2.5",
          "model": "kimi-k2.5",
          "family": "moonshot",
          "effort": "kimi-cli",
          "score": 25,
          "avgCostUsd": 2.13,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fswe-qwen3.6-plus",
          "model": "qwen3.6-plus",
          "family": "other",
          "effort": "qwen-code",
          "score": 21,
          "avgCostUsd": 1.75,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "estimated"
        }
      ],
      "provenance": {
        "url": "https://frontierswe.com",
        "capturedAt": "2026-07-16",
        "license": "Scores from FrontierSWE; costs are illustrative API/relative blends, not measured $/task."
      },
      "modeKind": "harness"
    },
    {
      "id": "terminal-bench-2.1",
      "title": "Terminal Bench 2.1",
      "sourceNote": "Illustrative mix of Snorkel-verified Terminal-Bench 2.1 rows + published GPT-5.6 figures. Costs are approximate proxies (DeepSWE overlap or API-tier estimates). Not live OpenCodex metering.",
      "updated": "2026-07-11",
      "taskCount": 89,
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Estimated cost (USD, illustrative)",
        "yLabel": "Accuracy",
        "xUnit": "USD",
        "yUnit": "%"
      },
      "rows": [
        {
          "id": "tb21-gpt-5.6-sol",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "effort": "xhigh",
          "score": 88.8,
          "avgCostUsd": 8.39,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-gpt-5.6-terra",
          "model": "gpt-5.6-terra",
          "family": "openai",
          "effort": "codex",
          "score": 78.4,
          "scoreCi": 1.3,
          "avgCostUsd": 4.95,
          "tags": [
            "workhorse",
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-gpt-5.6-luna",
          "model": "gpt-5.6-luna",
          "family": "openai",
          "effort": "codex",
          "score": 75.7,
          "scoreCi": 1.3,
          "avgCostUsd": 3.03,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-claude-fable-5-cc",
          "model": "claude-fable-5",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 83.8,
          "scoreCi": 1.2,
          "avgCostUsd": 21.63,
          "tags": [
            "planner",
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-gpt-5.5-codex",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "codex",
          "score": 83.1,
          "scoreCi": 1.1,
          "avgCostUsd": 7.23,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-grok-4.5-cursor",
          "model": "grok-4.5",
          "family": "xai",
          "effort": "cursor-cli",
          "score": 79.3,
          "scoreCi": 1.5,
          "avgCostUsd": 2.42,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-claude-opus-4.8",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 78.9,
          "scoreCi": 1.3,
          "avgCostUsd": 13.22,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-muse-spark-1.1",
          "model": "muse-spark-1.1",
          "family": "other",
          "effort": "mini-swe-agent",
          "score": 76.2,
          "scoreCi": 1.2,
          "avgCostUsd": 2.36,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-claude-sonnet-5",
          "model": "claude-sonnet-5",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 74.6,
          "scoreCi": 1.6,
          "avgCostUsd": 12,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-glm-5.2",
          "model": "glm-5.2",
          "family": "zhipu",
          "effort": "claude-code",
          "score": 82.7,
          "avgCostUsd": 3.92,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-claude-opus-4.7",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 68.9,
          "scoreCi": 1.4,
          "avgCostUsd": 10,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-gemini-3.1-pro",
          "model": "gemini-3.1-pro",
          "family": "google",
          "effort": "gemini-cli",
          "score": 65.8,
          "scoreCi": 1.7,
          "avgCostUsd": 9.48,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-kimi-k3",
          "model": "kimi-k3",
          "family": "moonshot",
          "score": 88.3,
          "avgCostUsd": 5,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-hy3",
          "model": "hy3",
          "family": "other",
          "score": 71.7,
          "avgCostUsd": 2,
          "tags": [
            "cheap-subagent",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "tb21-minimax-m3",
          "model": "minimax-m3",
          "family": "other",
          "score": 66,
          "avgCostUsd": 1.5,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "estimated"
        }
      ],
      "provenance": {
        "url": "https://www.tbench.ai",
        "capturedAt": "2026-07-11",
        "license": "Harness results from Terminal-Bench; costs are estimates / API blends."
      },
      "modeKind": "harness"
    },
    {
      "id": "program-bench",
      "title": "Program Bench",
      "sourceNote": "Snapshot from programbench.com extended leaderboard (mini-SWE-agent, 200 tasks). Score = almost-resolved (≥95% tests). Costs = published average API $/task. Not live OpenCodex metering.",
      "updated": "2026-06-23",
      "taskCount": 200,
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Avg cost per task (USD)",
        "yLabel": "Almost resolved",
        "xUnit": "USD",
        "yUnit": "%"
      },
      "rows": [
        {
          "id": "pb-gpt-5.5-xhigh",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "xhigh",
          "score": 13.5,
          "avgCostUsd": 8.85,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-gpt-5.5-high",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "high",
          "score": 5,
          "avgCostUsd": 3.65,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-claude-opus-4.7-xhigh",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "effort": "xhigh",
          "score": 4.5,
          "avgCostUsd": 10.96,
          "tags": [
            "planner",
            "frontier"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-claude-opus-4.7",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "score": 3,
          "avgCostUsd": 3.81,
          "tags": [
            "frontier"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-claude-opus-4.6",
          "model": "claude-opus-4.6",
          "family": "anthropic",
          "score": 2.5,
          "avgCostUsd": 11.38,
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-gpt-5.5-default",
          "model": "gpt-5.5",
          "family": "openai",
          "score": 1.5,
          "avgCostUsd": 1.21,
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-claude-sonnet-4.6",
          "model": "claude-sonnet-4.6",
          "family": "anthropic",
          "score": 1,
          "avgCostUsd": 26.73,
          "tags": [
            "workhorse"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-gpt-5.4",
          "model": "gpt-5.4",
          "family": "openai",
          "score": 0,
          "avgCostUsd": 0.33,
          "tags": [
            "fast",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-gemini-3.1-pro",
          "model": "gemini-3.1-pro",
          "family": "google",
          "score": 0,
          "avgCostUsd": 1.51,
          "tags": [
            "frontier"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-gemini-3-flash",
          "model": "gemini-3-flash",
          "family": "google",
          "score": 0,
          "avgCostUsd": 0.3,
          "tags": [
            "fast",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-claude-haiku-4.5",
          "model": "claude-haiku-4.5",
          "family": "anthropic",
          "score": 0,
          "avgCostUsd": 0.8,
          "tags": [
            "fast",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-gpt-5.4-mini",
          "model": "gpt-5.4-mini",
          "family": "openai",
          "score": 0,
          "avgCostUsd": 0.04,
          "tags": [
            "fast",
            "cheap-subagent"
          ],
          "costKind": "measured"
        },
        {
          "id": "pb-gpt-5-mini",
          "model": "gpt-5-mini",
          "family": "openai",
          "score": 0,
          "avgCostUsd": 0.03,
          "tags": [
            "fast",
            "cheap-subagent"
          ],
          "costKind": "measured"
        }
      ],
      "provenance": {
        "url": "https://programbench.com/extended/",
        "capturedAt": "2026-06-23",
        "license": "Scores and average API costs from ProgramBench extended leaderboard."
      },
      "modeKind": "effort"
    },
    {
      "id": "swe-marathon",
      "title": "SWE Marathon",
      "sourceNote": "Illustrative snapshot from swe-marathon.org (pass@1 on 20 ultra-long-horizon tasks). Cost axis is an estimated relative run cost (long trajectories); not published $/task. Not live OpenCodex metering.",
      "updated": "2026-07-09",
      "taskCount": 20,
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Est. relative run cost (USD)",
        "yLabel": "Resolution rate",
        "xUnit": "USD",
        "yUnit": "%"
      },
      "rows": [
        {
          "id": "swem-grok-4.5",
          "model": "grok-4.5",
          "family": "xai",
          "effort": "grok-build",
          "score": 29,
          "avgCostUsd": 8,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-claude-opus-4.8",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 26,
          "avgCostUsd": 28,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-claude-fable-5",
          "model": "claude-fable-5",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 24,
          "avgCostUsd": 45,
          "tags": [
            "planner",
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-claude-opus-4.7-cc",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "effort": "claude-code",
          "score": 16,
          "avgCostUsd": 22,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-glm-5.2",
          "model": "glm-5.2",
          "family": "zhipu",
          "effort": "claude-code",
          "score": 13,
          "avgCostUsd": 6,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-gpt-5.5-codex",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "codex",
          "score": 12,
          "avgCostUsd": 18,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-claude-opus-4.7-terminus",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "effort": "terminus-2",
          "score": 11,
          "avgCostUsd": 20,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-gemini-3.5-flash",
          "model": "gemini-3.5-flash",
          "family": "google",
          "effort": "gemini-cli",
          "score": 7,
          "avgCostUsd": 5,
          "tags": [
            "fast"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-gpt-5.5-terminus",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "terminus-2",
          "score": 6,
          "avgCostUsd": 16,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-gemini-3.1-pro-terminus",
          "model": "gemini-3.1-pro",
          "family": "google",
          "effort": "terminus-2",
          "score": 4,
          "avgCostUsd": 12,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-deepseek-v4-pro",
          "model": "deepseek-v4-pro",
          "family": "deepseek",
          "effort": "terminus-2",
          "score": 4,
          "avgCostUsd": 7,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-gemini-3.1-pro-cli",
          "model": "gemini-3.1-pro",
          "family": "google",
          "effort": "gemini-cli",
          "score": 2,
          "avgCostUsd": 12,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-glm-5.1",
          "model": "glm-5.1",
          "family": "zhipu",
          "effort": "terminus-2",
          "score": 1,
          "avgCostUsd": 5,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-minimax-m2.7",
          "model": "minimax-m2.7",
          "family": "other",
          "effort": "terminus-2",
          "score": 0,
          "avgCostUsd": 3,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "swem-kimi-k2.6",
          "model": "kimi-k2.6",
          "family": "moonshot",
          "effort": "kimi-cli",
          "score": 0,
          "avgCostUsd": 4,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "estimated"
        }
      ],
      "provenance": {
        "url": "https://swe-marathon.org",
        "capturedAt": "2026-07-09",
        "license": "Scores from SWE-Marathon; costs are estimates."
      },
      "modeKind": "harness"
    },
    {
      "id": "frontend-code-arena",
      "title": "Frontend Code Arena",
      "sourceNote": "Illustrative snapshot from arena.ai Code Arena | WebDev (Elo from blind human votes on frontend tasks, Jul 16 2026). X-axis is blended public API $/MTok — preference Elo has no $/task. Not live OpenCodex metering.",
      "updated": "2026-07-16",
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Relative API cost (blended $/MTok)",
        "yLabel": "Arena Elo",
        "xUnit": "USD",
        "yUnit": "index"
      },
      "rows": [
        {
          "id": "fca-kimi-k3",
          "model": "kimi-k3",
          "family": "moonshot",
          "score": 1679,
          "scoreCi": 17,
          "avgCostUsd": 9,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-fable-5",
          "model": "claude-fable-5",
          "family": "anthropic",
          "score": 1631,
          "scoreCi": 13,
          "avgCostUsd": 30,
          "tags": [
            "planner",
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-gpt-5.6-sol-xhigh",
          "model": "gpt-5.6-sol",
          "family": "openai",
          "effort": "xhigh",
          "score": 1618,
          "scoreCi": 13,
          "avgCostUsd": 17.5,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-glm-5.2-max",
          "model": "glm-5.2",
          "family": "zhipu",
          "effort": "max",
          "score": 1587,
          "scoreCi": 10,
          "avgCostUsd": 2.9,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-opus-4.8-thinking",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "effort": "thinking",
          "score": 1562,
          "scoreCi": 9,
          "avgCostUsd": 15,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-grok-4.5",
          "model": "grok-4.5",
          "family": "xai",
          "score": 1558,
          "scoreCi": 13,
          "avgCostUsd": 4,
          "tags": [
            "workhorse",
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-opus-4.7-thinking",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "effort": "thinking",
          "score": 1558,
          "scoreCi": 7,
          "avgCostUsd": 15,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-opus-4.7",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "score": 1555,
          "scoreCi": 7,
          "avgCostUsd": 15,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-opus-4.6-thinking",
          "model": "claude-opus-4.6",
          "family": "anthropic",
          "effort": "thinking",
          "score": 1542,
          "scoreCi": 6,
          "avgCostUsd": 15,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-sonnet-5-high",
          "model": "claude-sonnet-5",
          "family": "anthropic",
          "effort": "high",
          "score": 1542,
          "scoreCi": 12,
          "avgCostUsd": 6,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-opus-4.6",
          "model": "claude-opus-4.6",
          "family": "anthropic",
          "score": 1536,
          "scoreCi": 6,
          "avgCostUsd": 15,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-opus-4.8",
          "model": "claude-opus-4.8",
          "family": "anthropic",
          "score": 1534,
          "scoreCi": 9,
          "avgCostUsd": 15,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-glm-5.1",
          "model": "glm-5.1",
          "family": "zhipu",
          "score": 1526,
          "scoreCi": 9,
          "avgCostUsd": 2.9,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-claude-sonnet-4.6",
          "model": "claude-sonnet-4.6",
          "family": "anthropic",
          "score": 1522,
          "scoreCi": 6,
          "avgCostUsd": 9,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-kimi-k2.6",
          "model": "kimi-k2.6",
          "family": "moonshot",
          "score": 1515,
          "avgCostUsd": 2.13,
          "tags": [
            "cheap-subagent",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-gpt-5.5-xhigh",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "xhigh",
          "score": 1504,
          "scoreCi": 7,
          "avgCostUsd": 17.5,
          "tags": [
            "frontier",
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-gpt-5.5-high",
          "model": "gpt-5.5",
          "family": "openai",
          "effort": "high",
          "score": 1482,
          "scoreCi": 7,
          "avgCostUsd": 17.5,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "fca-gpt-5.4-high",
          "model": "gpt-5.4",
          "family": "openai",
          "effort": "high",
          "score": 1457,
          "scoreCi": 17,
          "avgCostUsd": 8.75,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        }
      ],
      "provenance": {
        "url": "https://arena.ai",
        "capturedAt": "2026-07-16",
        "license": "Elo/scores from Code Arena; costs are estimates."
      },
      "modeKind": "effort"
    },
    {
      "id": "cybench",
      "title": "Cybench",
      "sourceNote": "Illustrative snapshot from cybench.github.io (unguided % solved on professional CTF tasks). Many rows are system-card / subset results, not one uniform full-suite run. Costs are relative agent-run estimates — Cybench does not publish $/task. Not live OpenCodex metering.",
      "updated": "2026-07-16",
      "taskCount": 40,
      "axes": {
        "xKey": "avgCostUsd",
        "yKey": "score",
        "xLabel": "Est. relative run cost (USD)",
        "yLabel": "Unguided % solved",
        "xUnit": "USD",
        "yUnit": "%"
      },
      "rows": [
        {
          "id": "cyb-claude-mythos-preview",
          "model": "claude-mythos-preview",
          "family": "anthropic",
          "score": 100,
          "avgCostUsd": 18,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-claude-opus-4.7",
          "model": "claude-opus-4.7",
          "family": "anthropic",
          "score": 96,
          "avgCostUsd": 14,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-claude-opus-4.6",
          "model": "claude-opus-4.6",
          "family": "anthropic",
          "score": 93,
          "avgCostUsd": 12,
          "tags": [
            "frontier",
            "planner"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-claude-opus-4.5",
          "model": "claude-opus-4.5",
          "family": "anthropic",
          "score": 82,
          "avgCostUsd": 11,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-muse-spark",
          "model": "muse-spark",
          "family": "other",
          "score": 65.4,
          "avgCostUsd": 3.2,
          "tags": [
            "workhorse",
            "frontier"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-claude-sonnet-4.5",
          "model": "claude-sonnet-4.5",
          "family": "anthropic",
          "score": 60,
          "avgCostUsd": 5.5,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-grok-4",
          "model": "grok-4",
          "family": "xai",
          "score": 43,
          "avgCostUsd": 2.8,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-claude-opus-4.1",
          "model": "claude-opus-4.1",
          "family": "anthropic",
          "score": 42,
          "avgCostUsd": 9,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-grok-4.1-thinking",
          "model": "grok-4.1",
          "family": "xai",
          "effort": "thinking",
          "score": 39,
          "avgCostUsd": 2.4,
          "tags": [
            "workhorse",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-claude-opus-4",
          "model": "claude-opus-4",
          "family": "anthropic",
          "score": 38,
          "avgCostUsd": 8,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-claude-sonnet-4",
          "model": "claude-sonnet-4",
          "family": "anthropic",
          "score": 35,
          "avgCostUsd": 4.5,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-grok-4-fast",
          "model": "grok-4",
          "family": "xai",
          "effort": "fast",
          "score": 30,
          "avgCostUsd": 1.2,
          "tags": [
            "fast",
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-o3-mini",
          "model": "o3-mini",
          "family": "openai",
          "score": 22.5,
          "avgCostUsd": 1.5,
          "tags": [
            "cheap-subagent"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-claude-3.7-sonnet",
          "model": "claude-3.7-sonnet",
          "family": "anthropic",
          "score": 20,
          "avgCostUsd": 3.5,
          "tags": [
            "workhorse"
          ],
          "costKind": "estimated"
        },
        {
          "id": "cyb-gpt-4.5-preview",
          "model": "gpt-4.5-preview",
          "family": "openai",
          "score": 17.5,
          "avgCostUsd": 6,
          "tags": [
            "frontier"
          ],
          "costKind": "estimated"
        }
      ],
      "provenance": {
        "url": "https://cybench.github.io",
        "capturedAt": "2026-07-16",
        "license": "Please cite Cybench when republishing results (see cybench.github.io).",
        "citation": "Cybench — https://cybench.github.io"
      },
      "modeKind": "effort"
    }
  ]
}