Benchmark
OmniDocBench - Table TEDS
OCR benchmarks measure how well models read text in images and documents. Scores are percentages.
- Models measured
- 11
- Scores recorded
- 11
- Sources
- 1
- IQ weight
- 0.56
Score distribution
- 88.1%
- Mean
- 89.3%
- Median
- 4.4%
- Std dev
- 83%
- 10th percentile
- 92.8%
- 90th percentile
- 80.8%
- Min
This benchmark has a weight of 0.56 for chat models, measured over 5 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 | PaddleOCR-VL-1.6· OmniDocBench data | 94.8% |
| 2 | GLM OCR· OmniDocBench data | 92.8% |
| 3 | HunyuanOCR· OmniDocBench data | 91% |
| 4 | PaddleOCR VL· OmniDocBench data | 90.7% |
| 5 | Qianfan-OCR· OmniDocBench data | 90.5% |
| 6 | Gemini 3 Flash· OmniDocBench data | 89.3% |
| 7 | Gemini 3 Pro Preview· OmniDocBench data | 89.2% |
| 8 | DeepSeek OCR 2· OmniDocBench data | 83.9% |
| 9 | Qwen3 VL 235B· OmniDocBench data | 83.1% |
| 10 | GPT 5.2· OmniDocBench data | 83% |
| 11 | Kimi K2.5· OmniDocBench data | 80.8% |
Data credits: OmniDocBench (OpenDataLab), Apache-2.0 (licence)