MCP Atlas

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

In index
Agentic

Models38

Top result88.1%Muse Spark 1.1

Top-3 spread2.5 pts

YearN/A

Result and price

Results

38 results

#ModelResult

01

Muse Spark 1.188.1%

02

Claude Opus 585.8%

03STStep 5 Preview85.6%

04MAKimi K384.2%

05TEHy4 preview83.7%

06

Gemini 3.5 Flash83.6%

07

Claude Opus 4.882.2%

08OAOrnith-1.5-397B80.0%

09TMInkling-Small79.6%

10

Claude Opus 4.7 (Adaptive)77.3%

11

GLM-5.276.8%

12

Qwen3.7 Max76.4%

13MAKimi K2.7 Code76.0%

14

Muse Glimmer 30B75.5%

15

GPT-5.575.3%

16

MiniMax M374.2%

17TMInkling74.1%

18

DeepSeek V4 Pro 081373.6%

19

Qwen3.7 Plus73.2%

20

GLM-5.171.8%

21

GPT-5.470.6%

22OAOrnith-1.5-35B-A3B70.2%

23

DeepSeek V4 Flash 073169.0%

24INLing 3.0 Flash65.5%

25

Qwen3.6-35B-A3B62.8%

26UPSolar Pro 461.4%

27UPSolar Open 258.2%

28

GPT-5.4 mini57.7%

29

GPT-5.4 nano56.1%

30MAKimi K2.655.9%

31OAOrnith-1.5-9B54.2%

32

Qwen3.6 Plus48.2%

33INLLaDA2.2-flash46.2%

34

Qwen3.5 397B46.1%

35MELongCat-Flash-Lite-Sparse45.6%

36

Claude Opus 4.542.3%

37

GLM-531.1%

38MAKimi K2.529.5%