benchgap
Vision & documents

OmniDocBench 1.5 leaderboard

As of 2026-11-22, the highest measured score on OmniDocBench 1.5 is 92.1% by Qwen3.8 Max. 30 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Glimpse92.6%estimated ± 2.5 pp, low confidence
2Qwen3.8 Max92.1%measured
3Qwen3.8-Flash-Next91.7%estimated ± 1.8 pp, low confidence
4Kimi K391.6%estimated ± 2.2 pp, high confidence
5MiniMax M391.6%measured
6Qwen3.7 Plus91.4%measured
7Qwen3.8-Omni-Flash91.3%estimated ± 1.8 pp, medium confidence
8Inkling 2 Small91.2%estimated ± 2.2 pp, high confidence
9Seed 2.1 Pro91.2%estimated ± 2.2 pp, high confidence
10Qwen3.8-27B91.1%measured
11Step 3.7 Flash91.1%estimated ± 2.0 pp, medium confidence
12Qwen3.5 397B90.8%measured
13Qwen3-VL-235B-A22B Instruct90.7%estimated ± 1.5 pp, low confidence
14Seed 2.1 Turbo90.5%estimated ± 2.2 pp, high confidence
15Qwen3.6-27B90.3%estimated ± 1.5 pp, medium confidence
16Inkling 2 Mini90.2%estimated ± 2.2 pp, high confidence
17Muse Spark Open90.2%estimated ± 2.0 pp, medium confidence
18dots3-note Preview90.0%estimated ± 2.0 pp, medium confidence
19Gemini 3.1 Pro90.0%estimated ± 2.0 pp, medium confidence
20Qwen3.6 Plus89.9%estimated ± 2.2 pp, high confidence
21Qwen3.6-35B-A3B89.9%measured
22Muse Spark89.8%estimated ± 2.0 pp, medium confidence
23Kimi K2.689.8%estimated ± 2.2 pp, high confidence
24Qwen3.5-122B-A10B89.4%estimated ± 2.2 pp, high confidence
25Qwen3.5-27B89.4%estimated ± 2.2 pp, high confidence
26Kimi K2.588.8%measured
27Qwen3.5-35B-A3B88.8%estimated ± 2.2 pp, high confidence
28Gemini 3 Pro88.5%measured
29GPT-5.488.1%estimated ± 2.0 pp, medium confidence
30Ternary Bonsai 2 27B87.9%estimated ± 1.8 pp, medium confidence
31Claude Opus 4.587.7%measured
32Gemma 4 12B87.6%estimated ± 2.2 pp, high confidence
33LFM3-VL-2B87.6%estimated ± 1.8 pp, medium confidence
34Grok 4.2087.5%estimated ± 2.0 pp, medium confidence
35Claude Opus 4.687.4%estimated ± 2.5 pp, low confidence
36Nemotron 3.5 Super VL 120B A12B87.2%estimated ± 2.2 pp, high confidence
37Nemotron 3.5 Super VL 120B A12B NVFP487.1%estimated ± 2.2 pp, high confidence
38GPT-5.285.7%measured
39ZAYA1-VL-8B85.6%estimated ± 1.8 pp, low confidence
40LFM2.5-VL-3B85.5%estimated ± 1.8 pp, low confidence
41Muse Glimmer 30B75.8%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General