MMLU-Pro (Vals) leaderboard

Vals AI’s independent run of MMLU-Pro across fourteen academic subjects.

Models ranked by their published MMLU-Pro, Vals AI run value

Rank	Model	Provider	Accuracy
1	Claude Fable 5.1	Anthropic	92.4
2	Claude Opus 5	Anthropic	91.6
3	Claude Fable 5	Anthropic	91.5
4	Gemini 3.1 Pro	Google	91.0
5	Gemini 3.8 Flash	Google	90.2
6	Gemini 3.7 Flash	Google	90.1
7	Claude Opus 4.7	Anthropic	89.9
8	Claude Opus 4.8	Anthropic	89.6
9	Gemini 3.5 Flash	Google	89.5
10	Grok 4.6	xAI	89.4
11	Gemini 3.6 Flash	Google	89.3
11	Qwen3.7 Max	Alibaba	89.3
13	Grok 4.5	xAI	89.2
14	GPT-5.6 Sol	OpenAI	89.1
15	Muse Spark 1.1	Meta	88.7
16	Gemini 3 Flash	Google	88.6
16	Qwen3.8 Max	Alibaba	88.6
18	Muse Spark 1.2	Meta	88.3
19	GPT-5.5	OpenAI	88.1
20	Kimi K3	Moonshot AI	88.0
21	Qwen3.6 Plus	Alibaba	87.7
22	Kimi K2.6	Moonshot AI	87.6
23	Claude Sonnet 5	Anthropic	87.5
24	Claude Sonnet 4.6	Anthropic	87.3
25	DeepSeek V4 Pro 0813	DeepSeek	87.0
26	GLM-5.1	Z.AI	86.9
27	GLM-5.3	Z.AI	86.8
28	GLM-5.2	Z.AI	86.7
28	GPT-5.6 Terra	OpenAI	86.7
30	Grok 4.20	xAI	86.3
30	TMInkling	Thinking Machines Lab	86.3
32	DeepSeek V4 Flash 0731	DeepSeek	86.2
32	Gemini 3.1 Flash-Lite	Google	86.2
34	GLM-5.3-Flash	Z.AI	86.1
35	GPT-5.6 Luna	OpenAI	86.0
36	Gemini 3.5 Flash-Lite	Google	85.8
36	Grok 4.3	xAI	85.8
36	Nemotron 3 Ultra	NVIDIA	85.8
39	TMInkling-Small	Thinking Machines Lab	85.6
40	GPT-5.4 mini	OpenAI	84.6
40	MiMo-V2.5-Pro	Xiaomi	84.6
42	Qwen3.8-27B	Alibaba	84.3
43	MiniMax M3	MiniMax	84.2
44	MiMo-V2.5	Xiaomi	82.9
45	Ling 3.0 Flash	InclusionAI	82.0
46	MiniMax M2.7	MiniMax	80.4
47	Claude Haiku 4.5	Anthropic	78.7
48	GPT-5.4 nano	OpenAI	77.2
49	Mistral Medium 3.5 128B	Mistral	75.3
50	Laguna XS.2	Poolside	69.1
51	Laguna M.1	Poolside	68.8

Evidence key: Observed

Rows are ordered by the value Vals AI MMLU-Pro, Vals AI run leaderboard published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.