Visual Understanding on BLINK
69.86AccuracyGPT-5
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-5Size=-2025.10 | 69.86 | |
| GPT4oModel Category=Proprietary Model2025.09 | 68 | |
| Gemini 2.5 FlashSize=-2025.10 | 67.49 | |
| InternVL3.5Size=38B2025.10 | 62.65 | |
| Qwen2.5-VLSize=72B2025.10 | 61.91 | |
| GPT-4oSize=-2025.10 | 60.28 | |
| ThinkMorphSize=7B, think mode=true2025.10 | 60.07 | |
| InternVL3.5Size=8B2025.10 | 59.6 | |
| VisionZero-Qwen-7B (Real-World)Base Model=Qwen2.5-VL-7B, Variant=Real-World2025.09 | 57.5 | |
| VisionZero-Qwen-7B (CLEVR)Base Model=Qwen2.5-VL-7B, Variant=CLEVR2025.09 | 57.2 | |
| VisReflectPrefilling=Single, Tool-using=false2026.06 | 57.1 | |
| Qwen2.5-VLPrefilling=Single, Tool-using=false2026.06 | 56.5 | |
| DeepEyesPrefilling=Single, Cropping=No, Tool-using=true2026.06 | 56.5 | |
| VisionZero-Qwen-7B (Chart)Base Model=Qwen2.5-VL-7B, Variant=Chart2025.09 | 56.1 | |
| PixelReasonerPrefilling=Single, Tool-using=true2026.06 | 56.1 | |
| Qwen2.5-VLSize=7B2025.10 | 55.92 | |
| ViGaL-Snake+RotationBase Model=Qwen2.5-VL-7B2025.09 | 55.6 | |
| PAPOPrefilling=Single, Tool-using=true2026.06 | 55.6 | |
| LVRPrefilling=Single, Tool-using=true2026.06 | 55.3 | |
| Qwen2.5-VL-7B-InstructBase Model=Qwen2.5-VL-7B2025.09 | 55.2 | |
| Thyme-VLPrefilling=Single, Cropping=No, Tool-using=true2026.06 | 55.2 | |
| MM-Eureka-Qwen-7BBase Model=Qwen2.5-VL-7B2025.09 | 54 | |
| VLAA-Thinker-7BBase Model=Qwen2.5-VL-7B2025.09 | 53 | |
| OpenVLThinker-7BBase Model=Qwen2.5-VL-7B2025.09 | 49.9 | |
| R1-OneVision-7BBase Model=Qwen2.5-VL-7B2025.09 | 48.7 | |
| BagelSize=7B, think mode=true2025.10 | 47.66 | |
| Janus-proSize=7B2025.10 | 38.51 | |
| ChameleonSize=7B2025.10 | 16.52 |