MCP Atlas

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

In index
Agentic
Best in agentic

Models52

Top result89.2%Muse Spark Open

Top-3 spread3.4 pts

Year—

Result and price

38 of 52 results

Results

52 results

#ModelResult

01Muse Spark Open89.2%

02Muse Spark 1.188.1%

03Claude Opus 585.8%

04Step 5 Preview85.6%

05Inkling 2 Small85.0%

06ERNIE 5.5 Flash Preview84.6%

07Kimi K384.2%

08Hy4 preview83.7%

09Gemini 3.5 Flash83.6%

10LongCat-2.583.1%

11Ling 3.1 Flash83.0%

12Claude Opus 4.882.2%

13Beam 1.0 501B-A23B FP881.5%

14Beam 1.0 501B-A23B80.8%

15Beam 1.0 501B-A23B NVFP480.5%

16Ornith-1.5-397B80.0%

17Inkling-Small79.6%

18Claude Opus 4.7 (Adaptive)77.3%

19GLM-5.276.8%

20Qwen3.7 Max76.4%

21Kimi K2.7 Code76.0%

22Ling 3.0 Flash Sante75.6%

23Muse Glimmer 30B75.5%

24GPT-5.575.3%

25MiniMax M374.2%

26Inkling74.1%

27DeepSeek V4 Pro 081373.6%

28Qwen3.7 Plus73.2%

29GLM-5.171.8%

30Ripple-65B Preview71.7%

31GPT-5.470.6%

32Ornith-1.5-35B-A3B70.2%

33DeepSeek V4 Flash 073169.0%

34Ling 3.0 Flash65.5%

35Qwen3.6-35B-A3B62.8%

36Spark-X2.6-4B62.4%

37Solar Pro 461.4%

38Solar Open 258.2%

39GPT-5.4 mini57.7%

40GPT-5.4 nano56.1%

41Kimi K2.655.9%

42Nemotron 3.5 Bolt 74B-A9B BF1655.6%

43Ornith-1.5-9B54.2%

44Nemotron 3.5 Bolt 74B-A9B52.2%

45Qwen3.6 Plus48.2%

46LLaDA2.2-flash46.2%

47Qwen3.5 397B46.1%

48LongCat-Flash-Lite-Sparse45.6%

49Claude Opus 4.542.3%

50LFM3-1.7B32.6%

51GLM-531.1%

52Kimi K2.529.5%