Skip to content

MCP Atlas

Real tasks over 40+ MCP servers and 220 tools, three to six tool calls each.
Scale AI
Muse Spark 1.1
#1 right now
88.1%
Pass rate
34
Models ranked
Sep 23, 2026
Board published

Ranking

Best variant per model, as published. Higher is better.

  1. 188.1%
  2. 2
    Claude Fable 5.1
    as “Fable 5.1
    87.2%
  3. 3
    Claude Opus 5
    as “claude-opus-5 (xhigh)
    85.8%
  4. 4
    Qwen3.8 2.4T A95B
    as “Qwen3.8-2.4T-A95B (xHigh)
    84.5%
  5. 5
    GLM-5.3
    as “GLM 5.3
    84.2%
  6. 6
    Gemini 3.5 Flash
    as “Gemini 3.5 Flash (high)
    83.6%
  7. 783.3%
  8. 8
    Kimi K3
    as “kimi-k3 (max)
    82.3%
  9. 9
    Claude Opus 4.8
    as “claude-opus-4-8 (max)
    82.2%
  10. 982.2%
  11. 11
    GPT-5.6
    as “gpt-5.6 (sol)
    81.8%
  12. 12
    Inkling Small
    as “Inkling-small
    79.2%
  13. 13
    Claude Opus 4.7
    as “claude-opus-4-7 (max)
    79.1%
  14. 14
    Gemini 3.1 Pro Preview
    as “gemini-3.1-pro-preview (high)
    78.2%
  15. 15
    GLM-5p2
    as “glm-5p2
    77.8%
  16. 16
    Claude Opus 4.6
    as “claude-opus-4-6 (max)
    76.8%
  17. 17
    Inkling
    as “Inkling (xHigh)
    76.0%
  18. 18
    GLM-5p1
    as “glm-5p1
    75.6%
  19. 19
    GPT-5.5
    as “gpt-5.5 (xhigh)
    75.3%
  20. 20
    GPT-5.4
    as “gpt-5.4 (xhigh)
    70.6%
  21. 21
    Gemini 3 Pro
    as “gemini-3-pro-preview
    70.3%
  22. 22
    Claude Opus 4.5
    as “claude-opus-4-5 (high)
    69.8%
  23. 23
    Claude Sonnet 4.6
    as “claude-sonnet-4-6
    69.5%
  24. 24
    GPT-5.2
    as “gpt-5.2 (xhigh)
    67.6%
  25. 25
    Kimi K2P5
    as “kimi-k2p5
    64.4%

Rank over time

History builds with every daily capture. One capture so far.

Other agents and tools boards

Scores as published by Scale AI on the capture date (2026-09-24). Source: scale.com leaderboard page.

scale.com/leaderboard/mcp_atlas

Weekly: the AI leaderboards with a new #1, Saturday mornings.