Skip to content

MCP Atlas

A benchmark for tool-calling over Model Context Protocol integrations and external tools.

42 models

Higher is better.

  1. 1Muse Spark OpenMeta89.2
  2. 2Muse Spark 1.1Meta88.1
  3. 3Claude Opus 5Anthropic85.8
  4. 4Step 5 PreviewStepFun85.6
  5. 5Kimi K3Moonshot AI84.2
  6. 6Hy4 previewTencent83.7
  7. 7Gemini 3.5 FlashGoogle83.6
  8. 8Ling 3.1 FlashInclusionAI83.0
  9. 9Claude Opus 4.8Anthropic82.2
  10. 10Beam 1.0 501B-A23BReflection AI80.8
  11. 11Ornith-1.5-397BOrnith AI80.0
  12. 12Inkling-SmallThinking Machines Lab79.6
  13. 13Claude Opus 4.7 (Adaptive)Anthropic77.3
  14. 14GLM-5.2Z.AI76.8
  15. 15Qwen3.7 MaxAlibaba76.4
  16. 16Kimi K2.7 CodeMoonshot AI76.0
  17. 17Muse Glimmer 30BMeta75.5
  18. 18GPT-5.5OpenAI75.3
  19. 19MiniMax M3MiniMax74.2
  20. 20InklingThinking Machines Lab74.1
  21. 21DeepSeek V4 Pro 0813DeepSeek73.6
  22. 22Qwen3.7 PlusAlibaba73.2
  23. 23GLM-5.1Z.AI71.8
  24. 24GPT-5.4OpenAI70.6
  25. 25Ornith-1.5-35B-A3BOrnith AI70.2
  26. 26DeepSeek V4 Flash 0731DeepSeek69.0
  27. 27Ling 3.0 FlashInclusionAI65.5
  28. 28Qwen3.6-35B-A3BAlibaba62.8
  29. 29Solar Pro 4Upstage61.4
  30. 30Solar Open 2Upstage58.2
  31. 31GPT-5.4 miniOpenAI57.7
  32. 32GPT-5.4 nanoOpenAI56.1
  33. 33Kimi K2.6Moonshot AI55.9
  34. 34Ornith-1.5-9BOrnith AI54.2
  35. 35Nemotron 3.5 Bolt 74B-A9BNVIDIA52.2
  36. 36Qwen3.6 PlusAlibaba48.2
  37. 37LLaDA2.2-flashInclusionAI46.2
  38. 38Qwen3.5 397BAlibaba46.1
  39. 39LongCat-Flash-Lite-SparseMeituan45.6
  40. 40Claude Opus 4.5Anthropic42.3
Higher is better. Showing the top 40 of 42. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Interactive tool-calling evaluation
Tasks
Tool-integrated agent tasks
Authors
OpenAI (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.

All results

1Muse Spark Open Meta89.2
2Muse Spark 1.1 Meta88.1
3Claude Opus 5 Anthropic85.8
4Step 5 Preview StepFun85.6
5Kimi K3 Moonshot AI84.2
6Hy4 preview Tencent83.7
7Gemini 3.5 Flash Google83.6
8Ling 3.1 Flash InclusionAI83.0
9Claude Opus 4.8 Anthropic82.2
10Beam 1.0 501B-A23B Reflection AI80.8
11Ornith-1.5-397B Ornith AI80.0
12Inkling-Small Thinking Machines Lab79.6
13Claude Opus 4.7 (Adaptive) Anthropic77.3
14GLM-5.2 Z.AI76.8
15Qwen3.7 Max Alibaba76.4
16Kimi K2.7 Code Moonshot AI76.0
17Muse Glimmer 30B Meta75.5
18GPT-5.5 OpenAI75.3
19MiniMax M3 MiniMax74.2
20Inkling Thinking Machines Lab74.1
21DeepSeek V4 Pro 0813 DeepSeek73.6
22Qwen3.7 Plus Alibaba73.2
23GLM-5.1 Z.AI71.8
24GPT-5.4 OpenAI70.6
25Ornith-1.5-35B-A3B Ornith AI70.2
26DeepSeek V4 Flash 0731 DeepSeek69.0
27Ling 3.0 Flash InclusionAI65.5
28Qwen3.6-35B-A3B Alibaba62.8
29Solar Pro 4 Upstage61.4
30Solar Open 2 Upstage58.2
31GPT-5.4 mini OpenAI57.7
32GPT-5.4 nano OpenAI56.1
33Kimi K2.6 Moonshot AI55.9
34Ornith-1.5-9B Ornith AI54.2
35Nemotron 3.5 Bolt 74B-A9B NVIDIA52.2
36Qwen3.6 Plus Alibaba48.2
37LLaDA2.2-flash InclusionAI46.2
38Qwen3.5 397B Alibaba46.1
39LongCat-Flash-Lite-Sparse Meituan45.6
40Claude Opus 4.5 Anthropic42.3
41GLM-5 Z.AI31.1
42Kimi K2.5 Moonshot AI29.5