Fine-grained Perception on HR-Bench 4K (FSP, FCP, Over.)
84.8Overall ScoreBVS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BVSModel=Qwen3-VL-8B2026.07 | 84.8 | 96.8 | 72.8 | |
| BVS-Qwen3VL-8BParameters=8B, Backbone=Qwen3-VL2026.07 | 84.8 | 96.8 | 72.8 | |
| BVSModel=Qwen3-VL-4B2026.07 | 82.4 | 91.5 | 73.3 | |
| Qwen3-VL-8B-w toolParameters=8B, Tool use=true2026.07 | 82.3 | — | — | |
| ZoomEyeModel=Qwen3-VL-4B2026.07 | 82.1 | 91.3 | 72.8 | |
| ViCropModel=Qwen3-VL-4B2026.07 | 80.1 | 89.3 | 71 | |
| RAPModel=Qwen3-VL-4B2026.07 | 79.9 | 91 | 68.8 | |
| w/o reasoningModel=Qwen3-VL-4B2026.07 | 79.6 | 89 | 70.3 | |
| ZoomEyeModel=Qwen3-VL-8B2026.07 | 78.8 | 92.5 | 65 | |
| ViCropModel=Qwen3-VL-8B2026.07 | 78.5 | 93 | 64 | |
| w/o reasoningModel=Qwen3-VL-8B2026.07 | 77.9 | 92 | 63.8 | |
| BVSModel=InternVL3.5-8B2026.07 | 77.6 | 94.8 | 60.5 | |
| Mini-o3-7BParameters=7B2026.07 | 77.5 | — | — | |
| TreeVGR-7BParameters=7B2026.07 | 77.1 | 90.3 | 64 | |
| Thyme-7BParameters=7B2026.07 | 77 | 91 | 63 | |
| ThymeParam Size=7B2026.07 | 77 | 91 | 63 | |
| SaGe-7BParam Size=7B2026.07 | 76.5 | 94 | 59 | |
| Simple O3-7BParameters=7B2026.07 | 76.2 | — | — | |
| RAPModel=Qwen3-VL-8B2026.07 | 75.9 | 91.5 | 60.3 | |
| InternVL3Param Size=78B2026.07 | 75.5 | 84.5 | 66.5 | |
| DeepEyes-7BParameters=7B2026.07 | 75.1 | 91.3 | 59 | |
| DeepEyesParam Size=7B2026.07 | 75.1 | 91.3 | 59 | |
| ViCropModel=InternVL3.5-8B2026.07 | 74 | 89.5 | 58.5 | |
| Vision-R1Param Size=7B2026.07 | 74 | 88.5 | 59.5 | |
| Qwen2.5-VLParam Size=32B2026.07 | 73.4 | 87.5 | 59.3 | |
| Pixel-ReasonerParam Size=7B2026.07 | 72.9 | 86 | 60.3 | |
| SaGe-3BParam Size=3B2026.07 | 72 | 89 | 55 | |
| ZoomEyeModel=InternVL3.5-8B2026.07 | 71.3 | 84.3 | 58.3 | |
| InternVL3Param Size=8B2026.07 | 70.8 | 79.3 | 62.3 | |
| LVRParam Size=7B2026.07 | 69.6 | 79.3 | 60 | |
| RAPModel=InternVL3.5-8B2026.07 | 68.8 | 80 | 57.5 | |
| Qwen2.5-VLParam Size=7B2026.07 | 68.8 | 85.2 | 52.2 | |
| BVSModel=InternVL3.5-4B2026.07 | 67.9 | 83.5 | 52.3 | |
| GPT-5-MiniParam Size=-2026.07 | 66.3 | — | — | |
| LLaVA-OVParam Size=72B2026.07 | 66.3 | 76.5 | 56 | |
| Qwen2.5-VLParam Size=3B2026.07 | 66 | 82.3 | 49.8 | |
| GPT-4oParam Size=-2026.07 | 65 | 66.8 | 63.3 | |
| ViCropModel=InternVL3.5-4B2026.07 | 64.3 | 80.3 | 48.3 | |
| LLaVA-OVParam Size=7B2026.07 | 64.3 | 74.8 | 53.8 | |
| w/o reasoningModel=InternVL3.5-8B2026.07 | 63.6 | 70.5 | 56.8 | |
| RAPModel=InternVL3.5-4B2026.07 | 62.6 | 77.5 | 47.8 | |
| ZoomEyeModel=InternVL3.5-4B2026.07 | 62 | 78 | 46 | |
| w/o reasoningModel=InternVL3.5-4B2026.07 | 49.5 | 56.5 | 42.5 |