MCP Atlas

Real-world tool use through the Model Context Protocol. Scale AI.

Published by

Scale AI SEAL

Category

Agents & tools

Index weight

1.0

Models

30

Data as of

9 Sept 2026

Scale AI SEAL Leaderboards.

1. 1Muse Spark 1.188.1%
2. 2Claude Fable 5.187.2%
3. 3Claude Opus 585.8%
4. 4Gemini 3.5 Flash83.6%
5. 5Claude Fable 583.3%
6. 6Kimi K382.3%
7. 7Muse Spark82.2%
8. 8Claude Opus 4.882.2%
9. 9GPT-5.681.8%
10. 10Inkling-Small79.2%
11. 11Claude Opus 4.779.1%
12. 12Gemini 3.1 Pro78.2%
13. 13GLM-5.277.8%
14. 14Claude Opus 4.676.8%
15. 15GLM-5.175.6%

SearchAll providersAnthropicGoogleMetaMoonshot AIOpenAIThinking MachinesZhipu AIOpen weights30 of 30

#	Model	Score ↓	Run ↕	Index ↕	Published ↕
1	Muse Spark 1.1Meta	88.1%	Muse Spark 1.1	64.9	–
2	Claude Fable 5.1Anthropic	87.2%	–	70.0	–
3	Claude Opus 5Anthropic	85.8%	–	66.3	–
4	Gemini 3.5 FlashGoogle	83.6%	–	60.4	–
5	Claude Fable 5Anthropic	83.3%	Claude Fable 5	70.4	–
6	Kimi K3Moonshot AIopen	82.3%	–	62.7	–
7	Muse SparkMeta	82.2%	–	65.4	–
8	Claude Opus 4.8Anthropic	82.2%	–	59.4	–
9	GPT-5.6OpenAI	81.8%	–	–	–
10	TInkling-SmallThinking Machinesopen	79.2%	–	55.0	–
11	Claude Opus 4.7Anthropic	79.1%	–	59.3	–
12	Gemini 3.1 ProGoogle	78.2%	–	60.7	–
13	GLM-5.2Zhipu AIopen	77.8%	glm-5p2	57.4	–
14	Claude Opus 4.6Anthropic	76.8%	–	54.1	–
15	GLM-5.1Zhipu AIopen	75.6%	glm-5p1	60.1	–
16	GPT-5.5OpenAI	75.3%	–	63.7	–
17	GPT-5.4OpenAI	70.6%	–	61.3	–
18	Gemini 3 ProGoogle	70.3%	–	61.6	–
19	Claude Opus 4.5Anthropic	69.8%	–	56.6	–
20	Claude Sonnet 4.6Anthropic	69.5%	–	60.3	–
21	GPT-5.2OpenAI	67.6%	–	57.3	–
22	Kimi K2.5Moonshot AIopen	64.4%	kimi-k2p5	56.1	–
23	Gemini 3 FlashGoogle	62.0%	–	56.6	–
24	Claude Sonnet 4.5Anthropic	59.5%	–	52.8	–
25	GLM-4.7Zhipu AIopen	58.1%	glm-4p7	54.9	–
26	Gemini 3.1 Flash LiteGoogle	57.1%	–	49.1	–
27	GPT-5.4 miniOpenAI	56.7%	–	46.0	–
28	GPT-5.1OpenAI	50.1%	–	53.8	–
29	o3-proOpenAI	44.5%	–	54.1	–
30	Claude Haiku 4.5Anthropic	40.2%	–	47.9	–