Benchmark
OmniDocBench - TEDS
OCR benchmarks measure how well models read text in images and documents. Scores are percentages.
- Models measured
- 25
- Scores recorded
- 25
- Sources
- 1
- IQ weight
- 1.23
Score distribution
- 66.4%
- Mean
- 74%
- Median
- 19.4%
- Std dev
- 35%
- 10th percentile
- 87%
- 90th percentile
- 31.6%
- Min
This benchmark has a weight of 1.23 for chat models, measured over 23 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash· Nanonets IDP data | 87.7% |
| 2 | Claude Sonnet 4.6· Nanonets IDP data | 87.1% |
| 3 | Gemini 3 Pro Preview· Nanonets IDP data | 87% |
| 4 | Qwen3 VL 235B· Nanonets IDP data | 86.8% |
| 5 | Qwen3 VL Plus· Nanonets IDP data | 86.6% |
| 6 | GPT 5.2· Nanonets IDP data | 84.9% |
| 7 | Claude Opus 4.6· Nanonets IDP data | 84.4% |
| 8 | GPT 5.4· Nanonets IDP data | 81.3% |
| 9 | Gemini 3.1 Pro Preview· Nanonets IDP data | 80.9% |
| 10 | Claude Haiku 4.5· Nanonets IDP data | 77.1% |
| 11 | Mistral Small· Nanonets IDP data | 75.1% |
| 12 | GPT-5 mini· Nanonets IDP data | 74.6% |
| 13 | GPT 4.1· Nanonets IDP data | 74% |
| 14 | Qwen3.5 9B· Nanonets IDP data | 73.9% |
| 15 | Ministral 8B· Nanonets IDP data | 67.1% |
| 16 | Qwen VL OCR· Nanonets IDP data | 62.1% |
| 17 | GPT-5 nano· Nanonets IDP data | 61.2% |
| 18 | Qwen3.5 4B· Nanonets IDP data | 60.4% |
| 19 | Gemma 4 E4B-it· Nanonets IDP data | 48.9% |
| 20 | Qwen3.5 2B· Nanonets IDP data | 45.3% |
| 21 | Qwen3.5 0.8B· Nanonets IDP data | 37.9% |
| 22 | GLM OCR· Nanonets IDP data | 37.4% |
| 23 | Gemma 4 E2B-it· Nanonets IDP data | 35% |
| 24 | Pixtral 12B· Nanonets IDP data | 32.1% |
| 25 | Gemma 3 12B IT· Nanonets IDP data | 31.6% |