Spatial Reasoning on CV-Bench
92AccuracyGEMINI-3-PRO-PREVIEW
Evaluation results
| Method | Links | |
|---|---|---|
| GEMINI-3-PRO-PREVIEW2026.02 | 92 | |
| Mimo-EmbodiedParams=7B2026.06 | 88.8 | |
| Molmo2-ERAccess Mode=Open weights, Open data, Open code2026.05 | 87.8 | |
| RoboBrain2.0-7Bparameters=7B, zero-shot=true2025.12 | 86.44 | |
| GPT-5Access Mode=API call only2026.05 | 86.1 | |
| GEMINI-2.5-PRO2026.02 | 85.9 | |
| Gemini-2.5-Pro2026.04 | 85.9 | |
| Gemini 2.5 ProAccess Mode=API call only2026.05 | 85.9 | |
| GPT-5-miniAccess Mode=API call only2026.05 | 85.9 | |
| Qwen3-VL-4BAccess Mode=Open weights only2026.05 | 85.8 | |
| RoboBrain2.0Params=7B2026.06 | 85.8 | |
| RoboBrain-7B-2.0number of parameters=7B, version=2.02026.01 | 85.75 | |
| Qwen3-VL-8Bparameters=8B, zero-shot=true2025.12 | 85.69 | |
| Qwen3-VL-8BAccess Mode=Open weights only2026.05 | 85.6 | |
| VST-7B-SFT2026.02 | 85.5 | |
| RoboPINParams=4B2026.06 | 85.5 | |
| GPT-o4-mini-2025-05-162026.01 | 85.21 | |
| GPT-4o-minizero-shot=true2025.12 | 85.21 | |
| SEED-1.62026.02 | 85.2 | |
| Seed-1.62026.04 | 85.2 | |
| QWEN3-VL-8B-INSTRUCT2026.02 | 85.1 | |
| GeoThinker QWEN3VL-8BActive Perception=true2026.02 | 85.1 | |
| Qwen3-VL-8B-InstructSize=8B2026.04 | 85.1 | |
| Qwen3-VLParams=4B2026.06 | 85 | |
| GUIDE-9BSize=9B2026.04 | 84.9 | |
| GUIDE-5BSize=5B2026.04 | 84.7 | |
| GPT-52026.02 | 84.6 | |
| GPT-52026.04 | 84.6 | |
| Gemini-2.5-ProParams=-2026.06 | 84.6 | |
| Gemini-2.5-Pro-preview-05-062026.01 | 84.59 | |
| Gemini-2.5-prozero-shot=true2025.12 | 84.59 | |
| GPT-5Params=-2026.06 | 84.5 | |
| VST-3B-SFT2026.02 | 84.4 | |
| GR-ER 1.5 ThinkingAccess Mode=API call only2026.05 | 84.3 | |
| Gemini-2.5-Flash-preview-04-172026.01 | 84.03 | |
| RoboBrain-32B-2.0number of parameters=32B, version=2.02026.01 | 83.92 | |
| GenieReasonerparameters=3B, zero-shot=true2025.12 | 83.89 | |
| GR-ER 1.5Access Mode=API call only2026.05 | 83.6 | |
| Action-Sketcher-3Bnumber of parameters=3B2026.01 | 82.89 | |
| VG-LLM-8BActive Perception=false2026.02 | 82.7 | |
| VG-LLM-8BSize=8B2026.04 | 82.7 | |
| Embodied-R1Params=3B2026.06 | 82.7 | |
| Qwen2.5-VL-72B-Instructnumber of parameters=72B2026.01 | 82.68 | |
| InternVL-3.5-8Bparameters=8B, zero-shot=true2025.12 | 82.16 | |
| GeoThinker QWEN2.5VL-7BActive Perception=true2026.02 | 82.1 | |
| VG-LLM-8B*Active Perception=false, Training Setting=S1+S22026.02 | 81.7 | |
| VG-LLM-8B*Size=8B, Dataset Config=S1+S22026.04 | 81.7 | |
| Qwen2.5-VL-32B-Instructnumber of parameters=32B2026.01 | 81.59 | |
| InternVL3.5Params=8B2026.06 | 81.5 | |
| INTERNVL3-8B2026.02 | 81 | |
| VG-LLM-4BActive Perception=false2026.02 | 81 | |
| InternVL3-8BSize=8B2026.04 | 81 | |
| VG-LLM-4BSize=4B2026.04 | 81 | |
| InternVL3.5-8BAccess Mode=Open weights only2026.05 | 80.9 | |
| Qwen2.5-VL-7Bparameters=7B, zero-shot=true2025.12 | 80.78 | |
| InternVL3.5-4BAccess Mode=Open weights only2026.05 | 80.5 | |
| VeBrainParams=7B2026.06 | 79.7 | |
| Pelican-VLParams=7B2026.06 | 78.9 | |
| GPT-4o-2024-11-202026.01 | 78.63 | |
| VeBrain-8Bnumber of parameters=8B2026.01 | 78.57 | |
| Claude-Sonnet-4-2025-05-142026.01 | 78.43 | |
| QWEN3-VL-2B-INSTRUCT2026.02 | 78.4 | |
| Qwen2.5-VL-7B + SAGEModel=Qwen2.5-VL-7B + SAGE2026.05 | 77.2 | |
| CAMBRIAN-S-7B2026.02 | 76.9 | |
| INTERNVL3-2B2026.02 | 76.5 | |
| InternVL3-2BSize=2B2026.04 | 76.5 | |
| InternVL-2.5-8BModel=InternVL-2.5-8B2026.05 | 76.5 | |
| RoboBrain-7B-1.0number of parameters=7B, version=1.02026.01 | 76.22 | |
| BAGEL-7B-MOT2026.02 | 76 | |
| Bagel-7B-MOTSize=7B2026.04 | 76 | |
| Claude-Sonnet-4zero-shot=true2025.12 | 75.8 | |
| QWEN2.5-VL-7B-INSTRUCT2026.02 | 75.4 | |
| GPT-4oModel=GPT-4o2026.05 | 75.4 | |
| CAMBRIAN-S-3B2026.02 | 75.2 | |
| Cosmos-Reason1-7Bparameters=7B, zero-shot=true2025.12 | 75.2 | |
| Cambrian-S-3BSize=3B2026.04 | 75.2 | |
| Cosmos-Reason1-7Bnumber of parameters=7B2026.01 | 74.71 | |
| Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.05 | 74.2 | |
| SPATIALLADDER-3B2026.02 | 73.7 | |
| SpatialLadder-3BSize=3B2026.04 | 73.7 | |
| SpatialLadder-3BModel=SpatialLadder-3B2026.05 | 73.7 | |
| QWEN2.5-VL-3B-INSTRUCT2026.02 | 71.8 | |
| VLM-3R-7BActive Perception=false2026.02 | 71.8 | |
| VLM-3R-7BSize=7B2026.04 | 71.8 | |
| Qwen2.5-VL-3Bparameters=3B, zero-shot=true2025.12 | 71.29 | |
| LLaVA-OV-7BAccess Mode=Open weights only2026.05 | 70.3 | |
| Molmo2Access Mode=Open weights, Open data, Open code2026.05 | 64.4 | |
| Magma-8Bnumber of parameters=8B2026.01 | 60.98 | |
| LLaVA-OV-7BModel=LLaVA-OV-7B2026.05 | 58.3 |