Spatial & 3D Understanding on VLADBench
68.61ScoreXEmbodied (Best)
Evaluation results
| Method | Links | |
|---|---|---|
| XEmbodied (Best)Model Category=Our Model2026.04 | 68.61 | |
| Qwen3-VL-32BModel Category=Open-Source Models2026.04 | 61.64 | |
| Qwen2.5-VL-32BModel Category=Open-Source Models2026.04 | 61.33 | |
| Qwen3-VL-A3B-30BModel Category=Open-Source Models2026.04 | 59.81 | |
| PR1-Detection-3BModel Category=Spatial Models2026.04 | 56.08 | |
| GPT-4oModel Category=Proprietary Models2026.04 | 56 | |
| Cosmos-R1Model Category=Embodied Models2026.04 | 55.93 | |
| Mimo-EmbodiedModel Category=Embodied Models2026.04 | 55.82 | |
| Gemini-1.5Model Category=Proprietary Models2026.04 | 54.23 | |
| Qwen2.5-VL-7BModel Category=Open-Source Models2026.04 | 50.75 | |
| PR1-Counting-2BModel Category=Spatial Models2026.04 | 48.52 | |
| DriveMMModel Category=Embodied Models2026.04 | 47.45 | |
| UniVG-R1-7BModel Category=Spatial Models2026.04 | 46.51 | |
| PR1-OCR-2BModel Category=Spatial Models2026.04 | 44.65 | |
| PR1-Grounding-2BModel Category=Spatial Models2026.04 | 43.67 |