OmniDocBench 1.5 — leaderboard

Document-understanding benchmark covering OCR, layout parsing, tables, formulas, and information extraction across diverse document types.

Metric: Overall (self-reported). Source: benchmarklist.com. Status: saturated. 50 models tracked.

Top models

#ModelScore
1Qwen 3.7 Plus91.4
2Qwen 3.6 Plus91.2
3Gemini 3 Flash (Preview)90.37
4Gemini 3.1 Pro (Preview)90
5Qwen 3 VL 235B A22B89.15
6Gemini 2.5 Pro88.03
7Claude Opus 4.6 (Max)86.6
8GPT-5.285.75
9GPT-5.485.5
10GPT-4o75.02
11Gemma 4 E2B0.29
12Gemma 4 E4B0.18
13Gemma 4 12B0.16
14Gemma 4 26B A4B0.15
15Gemma 4 31B0.13

Interactive version: theaggregate.ai/benchmark?slug=omnidocbench-1-5 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.