Math

MMAnswerBench leaderboard

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

As of 2026-10-07, the highest measured score on MMAnswerBench is 91.0% by GLM-5.2. 72 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	GPT-5.6 Sol	100.0%	estimated ± 1.3 pp, low confidence
2	GPT-6 Astra	100.0%	estimated ± 1.3 pp, low confidence
3	GPT-5.6 Terra	97.4%	estimated ± 1.3 pp, low confidence
4	GPT-5.6 Luna	95.1%	estimated ± 1.3 pp, low confidence
5	GLM-5.2	91.0%	measured
6	GPT-5.5 Pro	90.9%	estimated ± 1.3 pp, low confidence
7	GPT-5.4 Pro	90.4%	estimated ± 1.3 pp, low confidence
8	GPT-5.5	89.9%	estimated ± 1.3 pp, low confidence
9	Claude Opus 4.8	89.0%	estimated ± 1.3 pp, low confidence
10	Qwen3.7 Max	88.8%	estimated ± 2.8 pp, low confidence
11	GPT-5.4	88.0%	estimated ± 1.3 pp, low confidence
12	DeepSeek V4 Pro 0813	87.8%	estimated ± 2.8 pp, low confidence
13	Beam	87.8%	estimated ± 1.3 pp, high confidence
14	DeepSeek V4 Flash 0731	87.6%	estimated ± 2.8 pp, low confidence
15	Claude Opus 4.7	87.1%	estimated ± 1.3 pp, low confidence
16	Claude Opus 4.6	87.1%	estimated ± 1.3 pp, low confidence
17	Qwen3.7 Plus	86.6%	estimated ± 2.8 pp, low confidence
18	A.X K2	86.5%	estimated ± 1.3 pp, high confidence
19	Inkling	86.5%	estimated ± 1.3 pp, high confidence
20	GPT-5.2	86.2%	estimated ± 1.3 pp, low confidence
21	Gemini 3 Pro	86.2%	estimated ± 1.3 pp, low confidence
22	Kimi K2.6	86.0%	measured
23	Gemini 3.1 Pro	85.7%	estimated ± 1.3 pp, low confidence
24	Muse Spark	85.2%	estimated ± 1.3 pp, low confidence
25	Gemini 3.5 Flash	85.2%	estimated ± 1.3 pp, low confidence
26	GPT-5.1	84.7%	estimated ± 1.3 pp, medium confidence
27	Ternary Bonsai 2 27B	84.2%	estimated ± 1.3 pp, high confidence
28	Claude Opus 4.5	84.0%	measured
29	Solar Open 2	84.0%	estimated ± 1.3 pp, high confidence
30	GLM-5.1	83.8%	measured
31	Qwen3.6 Plus	83.8%	measured
32	Claude Sonnet 4.6	83.8%	estimated ± 1.3 pp, medium confidence
33	Inkling-Small	83.6%	estimated ± 1.3 pp, high confidence
34	GPT-5.4 nano	83.3%	estimated ± 1.3 pp, medium confidence
35	o4-mini (high)	83.3%	estimated ± 1.3 pp, medium confidence
36	Solar Pro 4	83.3%	estimated ± 1.3 pp, high confidence
37	Claude Sonnet 4.5	82.9%	estimated ± 1.3 pp, medium confidence
38	Gemini 2.5 Flash	82.9%	estimated ± 1.3 pp, medium confidence
39	Gemini 2.5 Pro	82.9%	estimated ± 1.3 pp, medium confidence
40	Gemini 3 Flash	82.9%	estimated ± 1.3 pp, medium confidence
41	Qwen 3.6 Max (preview)	82.9%	estimated ± 1.3 pp, medium confidence
42	o1	82.8%	estimated ± 1.5 pp, low confidence
43	DeepSeek V3	82.8%	estimated ± 1.5 pp, low confidence
44	GPT-4.1 mini	82.8%	estimated ± 1.5 pp, low confidence
45	GPT-4.1 nano	82.8%	estimated ± 1.5 pp, low confidence
46	GPT-4o	82.8%	estimated ± 1.5 pp, low confidence
47	Llama 4 Maverick	82.8%	estimated ± 1.5 pp, low confidence
48	Llama 4 Scout	82.8%	estimated ± 1.5 pp, low confidence
49	GLM-5	82.5%	measured
50	GLM-4.6	82.4%	estimated ± 1.3 pp, medium confidence
51	DeepSeek V3.2	82.4%	estimated ± 1.3 pp, medium confidence
52	Claude Haiku 4.5	82.4%	estimated ± 1.3 pp, low confidence
53	Grok 4	82.4%	estimated ± 1.3 pp, low confidence
54	o3	82.4%	estimated ± 1.3 pp, low confidence
55	Qwen3.5 Plus	82.4%	estimated ± 1.3 pp, low confidence
56	GPT-5.4 mini	82.4%	estimated ± 1.3 pp, low confidence
57	Muse Glimmer 30B	82.3%	estimated ± 1.3 pp, high confidence
58	Claude 3.5 Sonnet	81.9%	estimated ± 1.3 pp, low confidence
59	GLM-4.7	81.9%	estimated ± 1.3 pp, low confidence
60	GPT-4.1	81.9%	estimated ± 1.3 pp, low confidence
61	Grok 3 [Beta]	81.9%	estimated ± 1.3 pp, low confidence
62	Kimi K2	81.9%	estimated ± 1.3 pp, low confidence
63	Qwen3 235B 2507 (Reasoning)	81.9%	estimated ± 1.3 pp, low confidence
64	Qwen3.5 Flash	81.9%	estimated ± 1.3 pp, low confidence
65	MAI-Thinking-1	81.9%	estimated ± 1.3 pp, high confidence
66	Kimi K2.5	81.8%	measured
67	Qwen3.5 397B	80.9%	measured
68	Qwen3.6-27B	80.8%	measured
69	Ling 3.0 Flash	79.7%	estimated ± 1.3 pp, high confidence
70	Granite 4.2 30B	79.4%	estimated ± 2.0 pp, low confidence
71	Qwen3.6-35B-A3B	78.9%	measured
72	K-EXAONE 2.0	78.2%	estimated ± 1.3 pp, medium confidence
73	Granite 4.2 8B	73.9%	estimated ± 2.0 pp, low confidence
74	ZAYA1-8B	73.2%	estimated ± 1.3 pp, medium confidence
75	MiniCPM5-2B	69.3%	estimated ± 1.3 pp, medium confidence
76	Granite 4.2 3B	67.3%	estimated ± 2.0 pp, low confidence
77	Gemma 4 12B	57.2%	estimated ± 1.3 pp, medium confidence
78	ZAYA1-74B-Preview	55.8%	estimated ± 1.3 pp, medium confidence
79	LongCat-Flash-Lite-Sparse	44.0%	estimated ± 1.3 pp, medium confidence
80	LFM2.5-8B-A1B	29.7%	estimated ± 1.3 pp, medium confidence
81	MiniCPM5-1B	22.5%	estimated ± 1.3 pp, medium confidence
82	LLaDA2.2-mini	18.8%	estimated ± 1.3 pp, medium confidence

MMAnswerBench leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General