Vision & documents

CC-OCR leaderboard

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

As of 2026-10-10, the highest measured score on CC-OCR is 82.0% by Qwen3.5 397B. 15 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Command A+	88.6%	estimated ± 4.5 pp, low confidence
2	Qwen3.5-35B-A3B	82.7%	estimated ± 4.5 pp, low confidence
3	MiniMax M3	82.5%	estimated ± 3.7 pp, medium confidence
4	Qwen3.7 Plus	82.2%	estimated ± 3.7 pp, medium confidence
5	Qwen3.5 397B	82.0%	measured
6	Qwen3.6-35B-A3B	81.9%	measured
7	Qwen3.5-27B	81.7%	estimated ± 4.5 pp, low confidence
8	Qwen3.8-27B	81.7%	estimated ± 3.7 pp, medium confidence
9	Qwen3.6-27B	81.2%	measured
10	Nemotron 3 Nano Omni 30B A3B	81.0%	estimated ± 0.7 pp, medium confidence
11	LFM2.5-VL-3B	80.1%	estimated ± 0.7 pp, medium confidence
12	Qwen3.5-122B-A10B	79.9%	estimated ± 4.5 pp, low confidence
13	Kimi K2.5	79.7%	measured
14	Qwen3.8 Max	79.6%	measured
15	Gemini 3 Pro	79.0%	measured
16	ZAYA1-VL-8B	78.9%	estimated ± 0.7 pp, medium confidence
17	Interfaze Beta	78.1%	estimated ± 0.7 pp, low confidence
18	Qwen3.6 Plus	77.4%	estimated ± 4.5 pp, low confidence
19	Claude Opus 4.5	76.1%	estimated ± 3.7 pp, medium confidence
20	GPT-5.2	70.3%	measured
21	LFM2.5-VL-450M	66.2%	estimated ± 4.4 pp, low confidence
22	Muse Glimmer 30B	58.9%	estimated ± 3.7 pp, low confidence

CC-OCR leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General