Scientific Reasoning on Lens
56.2AccuracyGemini2.5-Pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini2.5-ProModel size=-2025.05 | 56.2 | |
| Qwen2.5-VLModel size=72B2025.05 | 53.65 | |
| Qwen2.5-VLModel size=32B2025.05 | 51.66 | |
| GPT-4oModel size=-2025.05 | 51.14 | |
| QVQ-MaxModel size=72B2025.05 | 50.8 | |
| InternVL3Model size=78B2025.05 | 49.39 | |
| InternVL3Model size=38B2025.05 | 47.18 | |
| Qwen2.5-VLModel size=7B2025.05 | 46.28 | |
| InternVL3Model size=9B2025.05 | 44.69 | |
| Deepseek-VL2Model size=MoE 4.5B/27B2025.05 | 44.58 | |
| Gemma3Model size=12B2025.05 | 43.33 | |
| InternVL3Model size=2B2025.05 | 40.56 | |
| Qwen2.5-VLModel size=3B2025.05 | 40.33 | |
| Gemma3Model size=4B2025.05 | 39.53 | |
| Deepseek-VL2-tinyModel size=MoE 1B/3B2025.05 | 38.97 | |
| Kimi-VL-thinkingModel size=MoE 2.8B/16B2025.05 | 29.4 |