OCRBench ocrbench Leaderboard
Comprehensive OCR capability benchmark for multimodal models — text recognition, scene/handwritten/document text, key information extraction, and math-expression OCR. Final Score is out of 1000. · Metric: Final Score (higher is better)
| # | Model | Final Score | Paper |
|---|---|---|---|
| 1 | JT-VL-Chat-V3.0 | 950.00 | link |
| 2 | CongRong-v2.0 | 927.00 | link |
| 3 | Step-1o | 926.00 | link |
| 4 | MUG-U-7B | 911.00 | link |
| 5 | Ovis2-4B | 911.00 | link |
| 6 | InternVL2.5-78B-MPO | 909.00 | link |
| 7 | GLM-4v-Plus-20250111 | 908.00 | link |
| 8 | InternVL3-78B | 908.00 | link |
| 9 | InternVL2.5-26B-MPO | 905.00 | link |
| 10 | SAIL-VL-1.6-8B | 905.00 | link |
| 11 | SenseNova-V6-Pro | 895.00 | link |
| 12 | BailingMM-Pro-0120 | 894.00 | link |
| 13 | InternVL2.5-38B-MPO | 894.00 | link |
| 14 | Ovis2-34B | 894.00 | link |
| 15 | SenseNova | 894.00 | link |
| 16 | Ovis2-8B | 891.00 | link |
| 17 | SAIL-VL-1.5-2B | 891.00 | link |
| 18 | TeleMM | 891.00 | link |
| 19 | Ovis2-1B | 890.00 | link |
| 20 | WeThink-Qwen2.5VL-7B | 890.00 | link |
| 21 | MiniCPM-o-2.6 | 889.00 | link |
| 22 | SAIL-VL-1.5-8B | 889.00 | link |
| 23 | Qwen2.5-VL-7B | 888.00 | link |
| 24 | Qwen2-VL-72B | 888.00 | link |
| 25 | InternVL3-38B | 886.00 | link |
| 26 | Step-1.5V | 886.00 | link |
| 27 | SenseNova-V6-5-Pro | 885.00 | link |
| 28 | InternVL3-8B | 884.00 | link |
| 29 | InternVL2.5-8B-MPO | 882.00 | link |
| 30 | Qwen2.5-VL-72B | 882.00 | link |