Knowledge & reasoning

MMMLU leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on MMMLU is 90.9% by Interfaze Beta. 62 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	GPT-6 Astra	91.6%	estimated ± 1.1 pp, low confidence
2	Sakana Fugu	91.4%	estimated ± 1.1 pp, low confidence
3	Sakana Fugu-Ultra	91.4%	estimated ± 1.1 pp, low confidence
4	GPT-5.6 Sol	91.2%	estimated ± 1.1 pp, low confidence
5	Gemini 3.1 Pro	91.1%	estimated ± 1.1 pp, low confidence
6	Claude Opus 4.7 (Adaptive)	91.1%	estimated ± 1.1 pp, low confidence
7	Claude Opus 4.8	90.9%	estimated ± 1.1 pp, low confidence
8	GPT-5.5	90.9%	estimated ± 1.1 pp, low confidence
9	Kimi K3	90.9%	estimated ± 1.1 pp, low confidence
10	Step 5 Preview	90.9%	estimated ± 1.1 pp, low confidence
11	Interfaze Beta	90.9%	measured
12	GPT-5.6 Terra	90.8%	estimated ± 1.1 pp, low confidence
13	GPT-5.4	90.7%	estimated ± 1.1 pp, low confidence
14	Ornith-1.5-397B	90.7%	estimated ± 1.1 pp, low confidence
15	Gemini 3.5 Flash	90.7%	estimated ± 1.1 pp, low confidence
16	Qwen3.8 Max	90.7%	estimated ± 1.1 pp, low confidence
17	Pareto 26.10 Preview	90.6%	estimated ± 1.1 pp, medium confidence
18	GPT-5.6 Luna	90.6%	estimated ± 1.1 pp, medium confidence
19	Hy4 preview	90.6%	estimated ± 1.1 pp, medium confidence
20	Qwen3.8-Flash-Next	90.4%	estimated ± 1.1 pp, medium confidence
21	Qwen3.7 Max	90.3%	measured
22	GLM-5.2	90.2%	estimated ± 1.1 pp, medium confidence
23	Qwen3.8-Omni-Flash	90.2%	estimated ± 1.1 pp, medium confidence
24	DeepSeek V4.1 Flash	90.1%	estimated ± 1.1 pp, medium confidence
25	Kimi K2.6	90.0%	estimated ± 1.1 pp, medium confidence
26	Beam	90.0%	estimated ± 1.1 pp, medium confidence
27	DeepSeek V4 Pro 0813	89.9%	estimated ± 1.1 pp, medium confidence
28	Inkling-Small	89.6%	estimated ± 1.1 pp, medium confidence
29	Muse Spark	89.6%	estimated ± 1.1 pp, medium confidence
30	Claude Opus 4.6	89.5%	estimated ± 1.1 pp, medium confidence
31	Ornith-1.5-35B-A3B	89.5%	estimated ± 1.1 pp, medium confidence
32	Qwen3.8-27B	89.5%	estimated ± 1.1 pp, medium confidence
33	Solar Pro 4	89.5%	estimated ± 1.1 pp, medium confidence
34	Grok 4.20	89.3%	estimated ± 1.1 pp, medium confidence
35	DeepSeek V4 Flash 0731	89.1%	estimated ± 1.1 pp, medium confidence
36	Inkling	89.0%	estimated ± 1.1 pp, medium confidence
37	Qwen3.7 Plus	89.0%	measured
38	Kimi K2.5	88.9%	estimated ± 1.1 pp, medium confidence
39	Hy3 Preview	88.7%	estimated ± 1.1 pp, medium confidence
40	MiniMax M2.7	88.6%	estimated ± 1.1 pp, medium confidence
41	Nemotron 3 Ultra	88.6%	estimated ± 1.1 pp, medium confidence
42	Ornith-1.5-9B	88.4%	estimated ± 1.1 pp, medium confidence
43	Solar Open 2	88.3%	estimated ± 1.1 pp, medium confidence
44	GLM-5.1	88.3%	estimated ± 1.1 pp, medium confidence
45	GLM-5	88.2%	estimated ± 1.1 pp, medium confidence
46	Ternary Bonsai 2 27B	88.1%	estimated ± 1.1 pp, medium confidence
47	A.X K2	88.0%	estimated ± 1.1 pp, medium confidence
48	Ling 3.0 Flash	87.7%	estimated ± 1.1 pp, medium confidence
49	MAI-Thinking-1	87.3%	estimated ± 1.1 pp, medium confidence
50	Ling 3.0 Flash FP8	87.2%	estimated ± 1.1 pp, medium confidence
51	K-EXAONE 2.0	86.6%	measured
52	Mercury 2.5	83.9%	estimated ± 1.1 pp, medium confidence
53	Gemma 4 12B	83.4%	measured
54	Trinity-Large-Thinking	81.6%	estimated ± 1.1 pp, low confidence
55	Nemotron 3.5 Lightning 30B A3B NVFP4	80.9%	estimated ± 1.1 pp, low confidence
56	Nemotron 3 Nano Omni 30B A3B	77.2%	estimated ± 1.1 pp, low confidence
57	ZAYA1-8B	75.7%	estimated ± 1.1 pp, low confidence
58	MiniCPM5-2B	74.6%	estimated ± 1.1 pp, low confidence
59	LongCat-Flash-Lite-Sparse	73.6%	estimated ± 1.1 pp, low confidence
60	Trinity-Large-Preview	62.8%	estimated ± 1.1 pp, low confidence
61	Mellum2-12B-A2.5B-Thinking	49.8%	estimated ± 1.1 pp, low confidence
62	ZAYA1-74B-Preview	49.1%	estimated ± 1.1 pp, low confidence
63	LLaDA2.2-mini	17.8%	estimated ± 1.1 pp, low confidence
64	Soofi S 30B-A3B	15.9%	estimated ± 1.1 pp, low confidence
65	Mellum2-12B-A2.5B-Instruct	11.7%	estimated ± 1.1 pp, low confidence
66	MiniCPM5-1B	0.9%	estimated ± 1.1 pp, low confidence
67	LFM2.5-230M	0.8%	estimated ± 1.1 pp, low confidence

MMMLU leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General