Benchmark
OmniDocBench - Table TEDS-S
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
- 0.36
Score distribution
- 91.4%
- Mean
- 93.2%
- Median
- 3.9%
- Std dev
- 86.8%
- 10th percentile
- 95.4%
- 90th percentile
- 84%
- Min
This benchmark has a weight of 0.36 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 | 97.1% |
| 2 | GLM OCR· OmniDocBench data | 95.4% |
| 3 | PaddleOCR VL· OmniDocBench data | 93.7% |
| 4 | Gemini 3 Flash· OmniDocBench data | 93.5% |
| 5 | Qianfan-OCR· OmniDocBench data | 93.3% |
| 6 | HunyuanOCR· OmniDocBench data | 93.2% |
| 7 | Gemini 3 Pro Preview· OmniDocBench data | 93% |
| 8 | GPT 5.2· OmniDocBench data | 87.9% |
| 9 | DeepSeek OCR 2· OmniDocBench data | 87.8% |
| 10 | Qwen3 VL 235B· OmniDocBench data | 86.8% |
| 11 | Kimi K2.5· OmniDocBench data | 84% |
Data credits: OmniDocBench (OpenDataLab), Apache-2.0 (licence)