Spatial Reasoning on BLINK (Dep., Spa., All)
93Spa. ScoreInternVL2.5-78B
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| InternVL2.5-78BModel Category=General VLMs2026.05 | 93 | — | — | 82.3 | — | 29.9 | 54.1 | — | |
| GPT-5Params=-2026.06 | 90.2 | — | — | 82.3 | — | — | 45.8 | 77.3 | |
| InternVL2.5-8BModel Category=General VLMs2026.05 | 89.5 | — | — | 77.4 | — | 34.7 | 44.4 | — | |
| Qwen2.5-VL-72BModel Category=General VLMs2026.05 | 88.8 | — | — | 81.5 | — | 31.4 | 53.4 | — | |
| Qwen2.5-VL-7B (+ GASP Full Model)Base Model=Qwen2.5-VL-7B, Loss Function=Lcorr + Ldepth2026.05 | 88.8 | — | — | 80.7 | — | 35 | 53.4 | — | |
| GPT-o4-mini-2025-05-162026.01 | 88.11 | 79.03 | 83.57 | — | — | — | — | — | |
| Qwen2.5-VL-7B (+ GASP Correspondence)Base Model=Qwen2.5-VL-7B, Loss Function=Lcorr2026.05 | 88.1 | — | — | 79 | — | 34.6 | 54.9 | — | |
| RoboBrain-32B-2.0number of parameters=32B, version=2.02026.01 | 87.41 | 79.84 | 83.63 | — | — | — | — | — | |
| RoboPINParams=4B2026.06 | 87.4 | — | — | 87.1 | — | — | 72.9 | 89.5 | |
| Qwen2.5-VL-32BModel Category=General VLMs2026.05 | 86 | — | — | 75 | — | 31.1 | 44.3 | — | |
| Qwen2.5-VL-32B-Instructnumber of parameters=32B2026.01 | 85.31 | 77.42 | 81.37 | — | — | — | — | — | |
| Gemini-2.5-ProParams=-2026.06 | 85.3 | — | — | 86.3 | — | — | 46.6 | 78.5 | |
| Gemini-2.5-Pro-preview-05-062026.01 | 84.62 | 79.03 | 81.83 | — | — | — | — | — | |
| Qwen3-VLParams=4B2026.06 | 84.6 | — | — | 85.5 | — | — | 51.1 | 88.4 | |
| Pelican-VLParams=7B2026.06 | 84.6 | — | — | 75 | — | — | 47.3 | 63.3 | |
| VG-LLMModel Category=3D Spatial Reasoning VLMs, Training Data=Finetuning on 3D Related VQAs2026.05 | 84.3 | — | — | 77.2 | — | 59.2 | 50.8 | — | |
| InternVL2.5-4BModel Category=General VLMs2026.05 | 83.9 | — | — | 66.9 | — | 32.5 | 44.4 | — | |
| LLaVA-Onevision-7BModel Category=General VLMs2026.05 | 83.9 | — | — | 75 | — | 24.4 | 55.6 | — | |
| GPT-4o-2024-11-202026.01 | 83.22 | 72.58 | 77.9 | — | — | — | — | — | |
| RoboBrain-7B-2.0number of parameters=7B, version=2.02026.01 | 83.22 | 84.68 | 83.95 | — | — | — | — | — | |
| VeBrainParams=7B2026.06 | 83.2 | — | — | 67.7 | — | — | 46.6 | 48.8 | |
| RoboBrain-7BModel Category=Open-source models, Rank=4.42025.08 | 81.8 | — | 64.6 | 75.8 | 45.1 | — | — | — | |
| InternVL3.5Params=8B2026.06 | 81.8 | — | — | 78.2 | — | — | 47.4 | 72.7 | |
| VeBrain-8Bnumber of parameters=8B2026.01 | 81.12 | 78.23 | 79.68 | — | — | — | — | — | |
| Qwen2.5-VL-7B (+ DL3DV)Base Model=Qwen2.5-VL-7B, Training Data=SFT on DL3DV VQA dataset2026.05 | 81 | — | — | 78.1 | — | 25.3 | 42 | — | |
| Claude-Sonnet-4-2025-05-142026.01 | 80.42 | 75.81 | 78.12 | — | — | — | — | — | |
| Embodied-R1Model Category=Open-source models, Rank=2.12025.08 | 80.4 | — | 66.4 | 76.6 | 57.4 | — | — | — | |
| RoboBrain2.0Params=7B2026.06 | 80.4 | — | — | 80.7 | — | — | 48.1 | 46.5 | |
| Qwen2.5-VL-7B (Baseline)Base Model=Qwen2.5-VL-7B, Training Data=SFT on LLaVA-Video 178K2026.05 | 80.2 | — | — | 78.9 | — | 26.5 | 41.5 | — | |
| Qwen2.5VL-3BModel Category=Open-source models, Rank=5.62025.08 | 79.7 | — | 61.7 | 66.9 | 55.7 | — | — | — | |
| Embodied-R1 w/o CSModel Category=Open-source models, Rank=3.42025.08 | 79.7 | — | 64.1 | 72.6 | 56.6 | — | — | — | |
| Gemini-2.5-Flash-preview-04-172026.01 | 79.02 | 77.42 | 78.22 | — | — | — | — | — | |
| Qwen2.5-VL-72B-Instructnumber of parameters=72B2026.01 | 78.32 | 74.19 | 76.26 | — | — | — | — | — | |
| RoboBrain-7B-1.0number of parameters=7B, version=1.02026.01 | 78.32 | 75.81 | 77.07 | — | — | — | — | — | |
| FSD-13BModel Category=Open-source models, Rank=4.62025.08 | 78.3 | — | 60.5 | 70.2 | 46.7 | — | — | — | |
| Embodied-SFTModel Category=Open-source models, Rank=3.72025.08 | 78.3 | — | 66.3 | 81.5 | 54.9 | — | — | — | |
| LLaVA-Onevision-72BModel Category=General VLMs2026.05 | 78.3 | — | — | 78.2 | — | 44.6 | 53.4 | — | |
| Mimo-EmbodiedParams=7B2026.06 | 77.6 | — | — | 94.4 | — | — | 26.3 | 83.2 | |
| GPT-4oModel Category=General VLMs2026.05 | 76.9 | — | — | 64.5 | — | 28.5 | 60.2 | — | |
| Embodied-R1Params=3B2026.06 | 76.9 | — | — | 79.8 | — | — | 40.6 | 50 | |
| Cambrain-8BModel Category=General VLMs2026.05 | 74.8 | — | — | 51.6 | — | 26.2 | 36.8 | — | |
| Action-Sketcher-3Bnumber of parameters=3B2026.01 | 74.67 | 82.3 | 82 | — | — | — | — | — | |
| Cosmos-Reason1-7Bnumber of parameters=7B2026.01 | 73.43 | 63.71 | 68.57 | — | — | — | — | — | |
| GPT4VModel Category=Closed-source models2025.08 | 72.7 | — | 60.7 | 59.7 | 54.9 | — | — | — | |
| LLaVA-1.5-13BModel Category=Open-source models, Rank=9.42025.08 | 69.9 | — | 54.2 | 53.2 | 52.5 | — | — | — | |
| GPT4oModel Category=Closed-source models2025.08 | 69.2 | — | 65.9 | 74.2 | 59.8 | — | — | — | |
| Gemini-1.5-ProModel Category=General VLMs2026.05 | 67.1 | — | — | 50 | — | 34.6 | 41.3 | — | |
| Magma-8Bnumber of parameters=8B2026.01 | 66.43 | 65.32 | 65.88 | — | — | — | — | — | |
| SAT-Dynamic-13BModel Category=Open-source models, Rank=6.62025.08 | 66.4 | — | 57.5 | 73.4 | 45.9 | — | — | — | |
| RoboPoint-13BModel Category=Open-source models, Rank=7.22025.08 | 65.7 | — | 57.2 | 62.1 | 56.6 | — | — | — | |
| ASMv2-13BModel Category=Open-source models, Rank=7.22025.08 | 65 | — | 57.5 | 56.5 | 63.9 | — | — | — | |
| AoTDModel Category=3D Spatial Reasoning VLMs, Training Data=Finetuning on 3D Related VQAs2026.05 | 61.5 | — | — | 49.2 | — | 23.3 | 45.1 | — | |
| LLaVA-NeXT-Video-7B (+ DL3DV)Base Model=LLaVA-NeXT-Video-7B, Training Data=SFT on DL3DV VQA dataset2026.05 | 54.5 | — | — | 44 | — | 9.2 | 42.5 | — | |
| LLaVA-NeXT-Video-7B (Baseline)Base Model=LLaVA-NeXT-Video-7B, Training Data=SFT on LLaVA-Video 178K2026.05 | 53.1 | — | — | 44.4 | — | 11.5 | 42.1 | — | |
| LLaVA-NeXT-Video-7B (+ GASP Correspondence)Base Model=LLaVA-NeXT-Video-7B, Loss Function=Lcorr2026.05 | 49.2 | — | — | 46.7 | — | 17.6 | 44.4 | — | |
| VLM-3RModel Category=3D Spatial Reasoning VLMs, Training Data=Finetuning on 3D Related VQAs2026.05 | 48.3 | — | — | 47.6 | — | 40.1 | 50.1 | — | |
| LLaVA-NeXT-Video-7B (+ GASP Full Model)Base Model=LLaVA-NeXT-Video-7B, Loss Function=Lcorr + Ldepth2026.05 | 47.6 | — | — | 48.4 | — | 22 | 57.1 | — | |
| Skill-3D-4BModel=Qwen3-VL-4B2026.06 | — | — | — | — | — | — | 60.8 | — | |
| Skill-3D-8BModel=Qwen3-VL-8B2026.06 | — | — | — | — | — | — | 68.5 | — | |
| Think3D-4BModel=Qwen3-VL-4B2026.06 | — | — | — | — | — | — | 48.7 | — | |
| Think3D-8BModel=Qwen3-VL-8B2026.06 | — | — | — | — | — | — | 61.7 | — | |
| w/ ToolsModel=Qwen3-VL-4B2026.06 | — | — | — | — | — | — | 48.5 | — | |
| w/ ToolsModel=Qwen3-VL-8B2026.06 | — | — | — | — | — | — | 57.4 | — | |
| w/o ToolsModel=Qwen3-VL-4B2026.06 | — | — | — | — | — | — | 47.9 | — | |
| w/o ToolsModel=Qwen3-VL-8B2026.06 | — | — | — | — | — | — | 56.1 | — |