OCRBench-V2 (en): leaderboard
OCRBench v2 English subset: Enhanced benchmark for evaluating Large Multimodal Models on visual text localization and reasoning with English text content.
Metric: Score (self-reported). Source: benchmarklist.com. Status: saturation imminent. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.7 Plus | 70.7 |
| 2 | Qwen 3.6 Plus | 67 |
| 3 | Gemini 3.1 Pro (Preview) | 64.6 |
| 4 | Gemini 3 Pro (Preview) | 63.4 |
| 5 | Qwen3 Omni 30B A3B Instruct | 61.3 |
| 6 | Gemini 2.5 Pro | 59.3 |
| 7 | GPT-5.4 (xHigh) | 59.1 |
| 8 | Qwen 3.5 9B | 58.7 |
| 9 | Claude Opus 4.6 (Max) | 54.3 |
| 10 | GPT-5.2 | 50.5 |
| 11 | Molmo2-8B | 44.7 |
| 12 | Ministral 3 14B | 43.3 |
| 13 | Phi-4 Multimodal Instruct | 38.1 |
Interactive version: theaggregate.ai/benchmark?slug=ocrbench-v2-en · How It Works · Data refreshed daily, snapshot 2026-09-05.