Benchmark
OlmOCR Bench
OCR benchmarks measure how well models read text in images and documents. Scores are percentages. Higher is better.
- Models measured
- 33
- Scores recorded
- 33
- Sources
- 3
- IQ weight
- 0.88
Score distribution
- 69.1%
- Mean
- 75.3%
- Median
- 15.3%
- Std dev
- 47%
- 10th percentile
- 83.2%
- 90th percentile
- 20.6%
- Min
This benchmark has a weight of 0.88 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 | LightOnOCR-3 4B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 86.3% |
| 2 | Lightonocr 3 0.8b· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 85.5% |
| 3 | Jina OCR v1· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 83.4% |
| 4 | LightOnOCR-2-1B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 83.2% |
| 5 | GPT 5.4· Nanonets IDP data | 81% |
| 6 | PaddleOCR VL· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it. | 80% |
| 7 | Qianfan-OCR· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 79.8% |
| 8 | GPT 5.2· Nanonets IDP data | 79.1% |
| 9 | Qwen3 VL Plus· Nanonets IDP data | 77.9% |
| 10 | Gemini 3 Pro Preview· Nanonets IDP data | 77.7% |
| 11 | Qwen3.5 9B· Nanonets IDP data | 77.2% |
| 12 | Qwen3 VL 235B· Nanonets IDP data | 76.8% |
| 13 | DeepSeek OCR 2· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 76.3% |
| 14 | DeepSeek-OCR· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 75.7% |
| 15 | Qwen3.5 4B· Nanonets IDP data | 75.4% |
| 16 | Gemini 3 Flash· Nanonets IDP data | 75.3% |
| 17 | GLM OCR· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 3 sources
| 75.2% |
| 18 | Claude Opus 4.6· Nanonets IDP data | 74.1% |
| 19 | Claude Sonnet 4.6· Nanonets IDP data | 73.9% |
| 20 | Qwen3.5 2B· Nanonets IDP data | 71.9% |
| 21 | GPT-5 mini· Nanonets IDP data | 69.9% |
| 22 | Gemini 3.1 Pro Preview· Nanonets IDP data | 69.8% |
| 23 | Mistral Small· Nanonets IDP data | 69.6% |
| 24 | Qwen3.5 0.8B· Nanonets IDP data | 64.8% |
| 25 | Claude Haiku 4.5· Nanonets IDP data | 61.2% |
| 26 | Qwen VL OCR· Nanonets IDP data | 59% |
| 27 | Ministral 8B· Nanonets IDP data | 58.7% |
| 28 | GPT 4.1· Nanonets IDP data | 49.4% |
| 29 | Gemma 4 E4B-it· Nanonets IDP data | 47% |
| 30 | Pixtral 12B· Nanonets IDP data | 38.3% |
| 31 | Gemma 4 E2B-it· Nanonets IDP data | 38.2% |
| 32 | GPT-5 nano· Nanonets IDP data | 35.2% |
| 33 | Gemma 3 12B IT· Nanonets IDP data | 20.6% |
Data credits: Hugging Face official benchmark leaderboards; licence varies by benchmark · OpenEvals/leaderboard-data, MIT (licence)