Video OCR benchmark with 1,464 videos and 2,000 QA pairs across 25 tasks.
View on AlphaXiv ↗Overall accuracy across all video OCR tasks
Higher is better
| # | Model | Score | Source |
|---|---|---|---|
| ★ | gemini-25-pro | 73.7% | |
| 2 | qwen25-vl-72b | 69% | |
| 3 | internvl3-78b | 67.2% | |
| 4 | gpt-4o | 66.4% | |
| 5 | gemini-15-pro | 64.9% | |
| 6 | qwen25-vl-32b | 61% |