Vision & documents

SimpleVQA leaderboard

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

As of 2026-10-07, the highest measured score on SimpleVQA is 81.7% by Qwen3.7 Plus. 57 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Opus 4.5	99.0%	estimated ± 5.7 pp, low confidence
2	Holo2-4B	96.2%	estimated ± 5.7 pp, low confidence
3	Nemotron 3 Nano Omni 30B A3B	95.9%	estimated ± 5.7 pp, low confidence
4	Holo2-8B	95.5%	estimated ± 5.7 pp, low confidence
5	Qwen3.5 397B	91.7%	estimated ± 5.7 pp, low confidence
6	Holo2-30B-A3B	91.3%	estimated ± 5.7 pp, low confidence
7	Qwen3.6 Plus	89.7%	estimated ± 5.7 pp, low confidence
8	Holo2-235B-A22B	87.7%	estimated ± 5.7 pp, low confidence
9	Gemini 3 Pro	85.6%	estimated ± 5.7 pp, low confidence
10	Muse Glimmer 30B	82.7%	estimated ± 5.7 pp, low confidence
11	Qwen3.7 Plus	81.7%	measured
12	Step 3.7 Flash	79.2%	measured
13	Qwen3.8 Max	75.0%	measured
14	Seed 2.1 Turbo	74.4%	estimated ± 2.5 pp, low confidence
15	Seed 2.1 Pro	74.3%	estimated ± 2.5 pp, low confidence
16	Kimi K3	73.9%	estimated ± 2.5 pp, medium confidence
17	Claude Mythos 5	73.6%	estimated ± 7.8 pp, low confidence
18	Claude Opus 4.6	72.8%	estimated ± 5.7 pp, low confidence
19	dots3-note Preview	72.5%	measured
20	Gemini 3.1 Pro	72.4%	measured
21	Claude Opus 4.7 (Adaptive)	72.4%	estimated ± 7.8 pp, low confidence
22	Qwen3.8-Flash-Next	72.1%	estimated ± 7.4 pp, low confidence
23	GLM-5.3-Flash	71.5%	estimated ± 7.8 pp, low confidence
24	Muse Spark	71.3%	measured
25	Qwen3.8-Omni-Flash	71.2%	estimated ± 7.4 pp, low confidence
26	Gemini 3.7 Flash	71.2%	estimated ± 7.8 pp, low confidence
27	Muse Spark 1.1	71.0%	estimated ± 7.8 pp, low confidence
28	Claude Sonnet 5	71.0%	estimated ± 7.8 pp, low confidence
29	Sakana Fugu-Ultra	70.1%	estimated ± 7.8 pp, low confidence
30	GPT-5.6 Sol	69.8%	estimated ± 8.2 pp, medium confidence
31	Sakana Fugu	69.3%	estimated ± 7.8 pp, low confidence
32	Gemini 3.5 Flash	68.8%	estimated ± 7.8 pp, low confidence
33	GPT-5.5	68.8%	estimated ± 8.2 pp, medium confidence
34	GPT-5.6 Terra	68.5%	estimated ± 8.2 pp, medium confidence
35	GPT-5.2	67.7%	estimated ± 7.8 pp, low confidence
36	Inkling	67.6%	estimated ± 7.8 pp, low confidence
37	Qwen3.8-27B	67.6%	estimated ± 7.4 pp, low confidence
38	Kimi K2.5	67.3%	estimated ± 8.2 pp, medium confidence
39	Kimi K2.5 (Reasoning)	67.3%	estimated ± 8.2 pp, medium confidence
40	Inkling-Small	67.2%	estimated ± 7.8 pp, low confidence
41	GPT-5.6 Luna	67.2%	estimated ± 8.2 pp, medium confidence
42	MiMo-V2.5	67.1%	estimated ± 7.8 pp, low confidence
43	Grok 4.3	67.0%	estimated ± 8.2 pp, medium confidence
44	MiniMax M3	67.0%	estimated ± 8.2 pp, medium confidence
45	Pareto 26.9	67.0%	estimated ± 8.2 pp, medium confidence
46	Kimi K2.6	66.7%	estimated ± 7.8 pp, low confidence
47	Gemma 4 31B	66.3%	estimated ± 8.2 pp, medium confidence
48	GPT-5.4 mini	66.2%	estimated ± 8.2 pp, medium confidence
49	Step 5 Preview	65.8%	estimated ± 8.2 pp, medium confidence
50	Claude Opus 4.8	65.6%	estimated ± 5.7 pp, low confidence
51	Claude Sonnet 4.6	65.1%	estimated ± 7.8 pp, low confidence
52	Qwen3.5-122B-A10B	65.0%	estimated ± 7.8 pp, low confidence
53	Gemma 4 26B A4B	64.5%	estimated ± 8.2 pp, medium confidence
54	Interfaze Beta	62.9%	estimated ± 8.2 pp, medium confidence
55	Gemini 3.1 Flash-Lite	62.6%	estimated ± 7.8 pp, low confidence
56	Gemma 4 12B	61.7%	estimated ± 8.2 pp, medium confidence
57	GPT-5.4	61.1%	measured
58	GPT-5.4 nano	59.8%	estimated ± 8.2 pp, medium confidence
59	Qwen3.6-35B-A3B	58.9%	measured
60	GPT-6 Astra	58.1%	estimated ± 5.7 pp, low confidence
61	Grok 4.20	57.4%	measured
62	Qwen3.6-27B	56.1%	measured
63	Ternary Bonsai 2 27B	54.3%	estimated ± 9.5 pp, low confidence
64	Command A+	49.4%	estimated ± 7.8 pp, low confidence
65	LFM2.5-VL-3B	35.4%	measured
66	ZAYA1-VL-8B	22.9%	estimated ± 9.5 pp, low confidence
67	North Micro Vision Instruct	18.4%	estimated ± 9.5 pp, low confidence
68	LFM2.5-VL-450M	13.3%	estimated ± 9.5 pp, low confidence

SimpleVQA leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General