Benchmark
OCR Arena Latency (seconds)
OCR benchmarks measure how well models read text in images and documents. Scores are percentages. Lower is better.
- Models measured
- 11
- Scores recorded
- 17
- Sources
- 1
- IQ weight
- No weight published
Score distribution
- 26%
- Mean
- 20.5%
- Median
- 16.6%
- Std dev
- 13%
- 10th percentile
- 44.3%
- 90th percentile
- 11.3%
- Min
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | GPT 5.1low· OCR Arena data | 11.3% |
| 2 | GPT 5.2none· OCR Arena data | 12.5% |
| 3 | GPT 5low· OCR Arena data | 13% |
| 4 | GLM OCR· OCR Arena data | 14.3% |
| 5 | Gemini 2.5 Flash· OCR Arena data | 15.9% |
| 6 | Claude Opus 4.5low· OCR Arena data | 17.1% |
| 7 | Claude Opus 4.5medium· OCR Arena data | 17.6% |
| 8 | Claude Opus 4.6low· OCR Arena data | 19.9% |
| 9 | Claude Opus 4.6medium· OCR Arena data | 20.5% |
| 10 | GPT 5.1medium· OCR Arena data | 21.5% |
| 11 | Claude Opus 4.6adaptive· OCR Arena data | 24.6% |
| 12 | GPT 5medium· OCR Arena data | 27.2% |
| 13 | GPT 5.2medium· OCR Arena data | 27.5% |
| 14 | Gemini 3 Flash· OCR Arena data | 29.5% |
| 15 | DeepSeek-OCR· OCR Arena data | 44.3% |
| 16 | Gemini 2.5 Pro· OCR Arena data | 46% |
| 17 | Qwen: Qwen3 VL 8B Thinking· OCR Arena data | 79.9% |