Optical Character Recognition on Lens
68.51AccuracyQwen2.5-VL
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-VLModel size=72B2025.05 | 68.51 | |
| Qwen2.5-VLModel size=32B2025.05 | 65.64 | |
| Gemini2.5-ProModel size=-2025.05 | 61.95 | |
| Qwen2.5-VLModel size=7B2025.05 | 61.65 | |
| QVQ-MaxModel size=72B2025.05 | 58.67 | |
| InternVL3Model size=78B2025.05 | 54.21 | |
| Qwen2.5-VLModel size=3B2025.05 | 52.43 | |
| InternVL3Model size=38B2025.05 | 51.72 | |
| InternVL3Model size=9B2025.05 | 48.79 | |
| Deepseek-VL2Model size=MoE 4.5B/27B2025.05 | 48.76 | |
| InternVL3Model size=2B2025.05 | 45.51 | |
| Deepseek-VL2-tinyModel size=MoE 1B/3B2025.05 | 44.01 | |
| GPT-4oModel size=-2025.05 | 42.86 | |
| Gemma3Model size=12B2025.05 | 33.98 | |
| Kimi-VL-thinkingModel size=MoE 2.8B/16B2025.05 | 30.21 | |
| Gemma3Model size=4B2025.05 | 27.03 |