# Optical Character Recognition on OCRBench v2

Source: [Optical Character Recognition on OCRBench v2](<https://www.sota2.com/research/sota/optical-character-recognition-on-ocrbench-v2>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-12T08:04:19.097Z

Evaluation of Optical Character Recognition performance on the OCRBench v2 dataset, reporting scores segmented by language, such as English (en) and Chinese (zh).

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Optical Character Recognition |
| Dataset (including any stated split) | OCRBench v2 |
| Metric | Accuracy (en) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 22 of 22 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Accuracy (en)](<https://www.sota2.com/research/sota/optical-character-recognition-on-ocrbench-v2.md?metric=Accuracy%20(en)>): Higher is better; unit: PERCENT.

- [Accuracy (zh)](<https://www.sota2.com/research/sota/optical-character-recognition-on-ocrbench-v2.md?metric=Accuracy%20(zh)>): Higher is better; unit: PERCENT.

- [Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-ocrbench-v2.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy (en) | Result date | Report paper |
| --- | --- | --- | --- |
| Gemini-3-Pro | 68 | Not recorded | 69c60ef50636e5bf4afc8453 |
| Qwen3-VL-235B-Thinking | 66.8 | Not recorded | 69c60ef50636e5bf4afc8453 |
| Kimi-K2.5 | 64.2 | Not recorded | 69c60ef50636e5bf4afc8453 |
| Qwen3-VL-4B | 60.68 | Not recorded | 69b8e00a0636e5bf4afbea02 |
| Intern-S1-Pro | 60.1 | Not recorded | 69c60ef50636e5bf4afc8453 |
| GPT-5.2 | 56.4 | Not recorded | 69c60ef50636e5bf4afc8453 |
| Qianfan-OCR | 56 | Not recorded | 69b8e00a0636e5bf4afbea02 |
| Qwen3-VL-2B | 54.18 | Not recorded | 69b8e00a0636e5bf4afbea02 |
| PointsReader | 23.4 | Not recorded | 69b8e00a0636e5bf4afbea02 |
| MonkeyOCR | 21.78 | Not recorded | 69b8e00a0636e5bf4afbea02 |
| Dotsocr | 19.59 | Not recorded | 69b8e00a0636e5bf4afbea02 |
| PaddleOCR-VL | 18.15 | Not recorded | 69b8e00a0636e5bf4afbea02 |
| MinerU2.5 | 17.09 | Not recorded | 69b8e00a0636e5bf4afbea02 |
| DeepSeek-OCR | 15.98 | Not recorded | 69b8e00a0636e5bf4afbea02 |

### Gemini-3-Pro

Accuracy (en): 68. Result ID: 69c6906147f34ce924b0dd48.

[Reporting paper](<https://www.sota2.com/research/paper/69c60ef50636e5bf4afc8453.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Qwen3-VL-235B-Thinking

Accuracy (en): 66.8. Result ID: 69c6906147f34ce924b0dd45.

[Reporting paper](<https://www.sota2.com/research/paper/69c60ef50636e5bf4afc8453.md>)

Recorded evaluation configuration: {"Number of Parameters":"235B-A22B","Thinking Configuration":"true"}. This record does not assert independent reproduction.

### Kimi-K2.5

Accuracy (en): 64.2. Result ID: 69c6906147f34ce924b0dd46.

[Reporting paper](<https://www.sota2.com/research/paper/69c60ef50636e5bf4afc8453.md>)

Recorded evaluation configuration: {"Number of Parameters":"1T-A32B"}. This record does not assert independent reproduction.

### Qwen3-VL-4B

Accuracy (en): 60.68. Result ID: 69b9095b47f34ce924a15967.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Parameters":"4B"}. This record does not assert independent reproduction.

### Intern-S1-Pro

Accuracy (en): 60.1. Result ID: 69c6906147f34ce924b0dd44.

[Reporting paper](<https://www.sota2.com/research/paper/69c60ef50636e5bf4afc8453.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c60ef50636e5bf4afc8453.md>)

Recorded evaluation configuration: {"Number of Parameters":"1T-A22B"}. This record does not assert independent reproduction.

### GPT-5.2

Accuracy (en): 56.4. Result ID: 69c6906147f34ce924b0dd47.

[Reporting paper](<https://www.sota2.com/research/paper/69c60ef50636e5bf4afc8453.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Qianfan-OCR

Accuracy (en): 56. Result ID: 69b9095b47f34ce924a15960.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

[Implementation](<https://github.com/baidubce/Qianfan-VL>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Qwen3-VL-2B

Accuracy (en): 54.18. Result ID: 69b9095b47f34ce924a15968.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Parameters":"2B"}. This record does not assert independent reproduction.

### PointsReader

Accuracy (en): 23.4. Result ID: 69b9095b47f34ce924a15965.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### MonkeyOCR

Accuracy (en): 21.78. Result ID: 69b9095b47f34ce924a15964.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Dotsocr

Accuracy (en): 19.59. Result ID: 69b9095b47f34ce924a15963.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### PaddleOCR-VL

Accuracy (en): 18.15. Result ID: 69b9095b47f34ce924a15961.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

[Method origin paper](<https://www.sota2.com/research/paper/697c3b70a8b01e76cc8da769.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### MinerU2.5

Accuracy (en): 17.09. Result ID: 69b9095b47f34ce924a15962.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### DeepSeek-OCR

Accuracy (en): 15.98. Result ID: 69b9095b47f34ce924a15966.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e00a0636e5bf4afbea02.md>)

[Method origin paper](<https://www.sota2.com/research/paper/697ae9b8a2b28e761d87d127.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-OCR-2>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
