Benchmark
OCRBench
What is OCRBench?
OCRBench: Comprehensive evaluation benchmark for assessing Optical Character Recognition (OCR) capabilities in Large Multimodal Models across text recognition…
- Released
- 2024-01-17
- Modality
- multimodal
- Evaluates
- image to text, vision
- Openness
- unknown
- Importer
- LLM Stats
- Reported scores
- 24
Source provenance
- Original evidence https://api.zeroeval.com/leaderboard/benchmarks/ocrbench?top_n=500
OCRBench paper and code
OCRBench results and reported scores
LLM Stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| DeepSeek VL2 | DeepSeek | 0.811 | 2024-12-13 | evidence |
| DeepSeek VL2 Small | DeepSeek | 0.834 | 2024-12-13 | evidence |
| DeepSeek VL2 Tiny | DeepSeek | 0.809 | 2024-12-13 | evidence |
| Kimi K2.5 | Moonshot AI | 0.923 | 2026-01-27 | evidence |
| LFM2.5-VL-3B | Liquid AI | 0.842 | 2026-08-12 | evidence |
| North Micro Vision Instruct | Cohere | 0.792 | 2026-08-12 | evidence |
| Phi-4-multimodal-instruct | Microsoft | 0.844 | 2025-02-01 | evidence |
| Qwen2-VL-72B-Instruct | Qwen | 0.877 | 2024-08-29 | evidence |
| Qwen2.5 VL 72B Instruct | Qwen | 0.885 | 2025-01-26 | evidence |
| Qwen2.5 VL 7B Instruct | Qwen | 0.864 | 2025-01-26 | evidence |
| Qwen3 VL 235B A22B Instruct | Qwen | 0.92 | 2025-09-22 | evidence |
| Qwen3 VL 235B A22B Thinking | Qwen | 0.875 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Instruct | Qwen | 0.903 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Thinking | Qwen | 0.839 | 2025-09-22 | evidence |
| Qwen3 VL 32B Instruct | Qwen | 0.895 | 2025-09-22 | evidence |
| Qwen3 VL 32B Thinking | Qwen | 0.855 | 2025-09-22 | evidence |
| Qwen3 VL 4B Instruct | Qwen | 0.881 | 2025-09-22 | evidence |
| Qwen3 VL 4B Thinking | Qwen | 0.808 | 2025-09-22 | evidence |
| Qwen3 VL 8B Instruct | Qwen | 0.896 | 2025-09-22 | evidence |
| Qwen3 VL 8B Thinking | Qwen | 0.819 | 2025-09-22 | evidence |
| Qwen3.5-122B-A10B | Qwen | 0.921 | 2026-02-24 | evidence |
| Qwen3.5-27B | Qwen | 0.894 | 2026-02-24 | evidence |
| Qwen3.5-35B-A3B | Qwen | 0.91 | 2026-02-24 | evidence |
| Qwen3.6-27B | Qwen | 0.894 | 2026-04-21 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.