Spatial Reasoning on EmbSpatial
90.3Overall AccuracyHuman
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HumanCategory=Reference2026.05 | 90.3 | — | |
| Gemini 3 ProCategory=Proprietary Models2026.05 | 84.3 | — | |
| OpenVLThinkerV22026.04 | 83.1 | — | |
| RoboPINParams=4B2026.06 | 83.1 | — | |
| GPT-52026.04 | 82.9 | — | |
| GPT-5Access Mode=API call only2026.05 | 81.5 | — | |
| Ours-DTRReasoning Mode=detect-then-reason inference2026.06 | 81.3 | — | |
| TIGeR2025.10 | 80.82 | — | |
| OneThinker-8BParameters=8B2026.04 | 79.9 | — | |
| Qwen3-VL GRPOOptimization=GRPO2026.04 | 79.9 | — | |
| Qwen3-VL GDPOOptimization=GDPO2026.04 | 79.9 | — | |
| ProSRDescription=QWEN3-VL-8B-THINKING + Ours2026.05 | 79.8 | — | |
| GPT-3.2Category=Proprietary Models2026.05 | 79.3 | — | |
| Ours-LORReasoning Mode=language-only reasoning2026.06 | 79.2 | — | |
| Gemini 2.5 Pro2026.04 | 79.1 | — | |
| QWEN3-VL-8B-THINKING + GRPOMethod Variant=Vanilla GRPO2026.05 | 79 | — | |
| Qwen3-VL-8BModel Scale=8B2026.06 | 79 | — | |
| Qwen3-VL-8B-ThinkingCategory=Open-source General Models2026.05 | 78.9 | — | |
| GPT-5-miniAccess Mode=API call only2026.05 | 78.8 | — | |
| Molmo2-ERAccess Mode=Open weights, Open data, Open code2026.05 | 78.8 | — | |
| Gemini 2.5 ProCategory=Proprietary Models2026.05 | 78.8 | — | |
| GeoThinkerCategory=Open-source Spatial Intelligence Models2026.05 | 78.8 | — | |
| Gemini-2.5-Pro-preview-05-062026.01 | 78.74 | — | |
| Gemini-2.5-prozero-shot=true2025.12 | 78.74 | — | |
| Gemini2.5-proAccess=Closed-source2026.03 | 78.7 | — | |
| Gemini-2.5-ProParams=-2026.06 | 78.7 | — | |
| QWEN3-VL-8B-THINKING + SFTMethod Variant=SFT Only2026.05 | 78.6 | — | |
| RoboBrain-32B-2.0number of parameters=32B, version=2.02026.01 | 78.57 | — | |
| Qwen3-VL-8B-Inst.Access=Open-source general-purpose, Parameters=8B2026.03 | 78.5 | — | |
| GR-ER 1.5 ThinkingAccess Mode=API call only2026.05 | 78.4 | — | |
| Qwen3-VL-8Bparameters=8B, zero-shot=true2025.12 | 78.32 | — | |
| GPT-5Params=-2026.06 | 78.3 | — | |
| GPT-o4-mini-2025-05-162026.01 | 78.29 | — | |
| GPT-4o-minizero-shot=true2025.12 | 78.29 | — | |
| Qwen3-VL-InstructMode=Instruct2026.04 | 78.2 | — | |
| Qwen3-VL-8BAccess Mode=Open weights only2026.05 | 78.2 | — | |
| Qwen3-VL-4BAccess Mode=Open weights only2026.05 | 78.1 | — | |
| Gemini 2.5 ProAccess Mode=API call only2026.05 | 78 | — | |
| Qwen3-VL-8B-InstructCategory=Open-source General Models2026.05 | 77.7 | — | |
| Qwen3-VL-8BModel Parameters=8B2026.06 | 77.6 | — | |
| Ouro-Spatial-8BModel Parameters=8B2026.06 | 77.6 | — | |
| GeoWeaver-5BScale=5B2026.05 | 77.5 | — | |
| Robix 7B Baseparameters=7B, type=Base2025.12 | 77.4 | — | |
| ACE-Brain-0-8BAccess=Embodied Brain MLLMs, Parameters=8B2026.03 | 77.3 | — | |
| Ouro-Spatial-4BModel Parameters=4B2026.06 | 77.3 | — | |
| Qwen3-VLParams=4B2026.06 | 77.1 | — | |
| Qwen3-VL-4BModel Parameters=4B2026.06 | 76.9 | — | |
| Gemini-2.5-Pro2025.10 | 76.67 | — | |
| GLM-4.1V-Thinking2025.10 | 76.62 | — | |
| RoboBrain-7B-2.0number of parameters=7B, version=2.02026.01 | 76.32 | — | |
| RoboBrain2.0-7BAccess=Embodied Brain MLLMs, Parameters=7B2026.03 | 76.3 | — | |
| RoboBrain2.0Params=7B2026.06 | 76.3 | — | |
| MiMo-Embodied-7BAccess=Embodied Brain MLLMs, Parameters=7B2026.03 | 76.2 | — | |
| Mimo-EmbodiedParams=7B2026.06 | 76.2 | — | |
| RoboBrain-7B-2.0parameters=7B2025.12 | 75.88 | — | |
| RoboBrain2.0-7Bparameters=7B, zero-shot=true2025.12 | 75.8 | — | |
| InternVL3.5-8BCategory=Open-source General Models2026.05 | 75.7 | — | |
| Lumo-1-Stage1training_stage=Stage 12025.12 | 75.6 | — | |
| RoboBrain2.5-8BAccess=Embodied Brain MLLMs, Parameters=8B2026.03 | 75.6 | — | |
| Seed 1.6Category=Proprietary Models2026.05 | 75.4 | — | |
| Vlaser-8BAccess=Embodied Brain MLLMs, Parameters=8B2026.03 | 75.3 | — | |
| InternVL-3.5-8Bparameters=8B, zero-shot=true2025.12 | 75.11 | — | |
| Gemini-2.5-Flash-preview-04-172026.01 | 74.75 | — | |
| InternVL3.5-8BAccess Mode=Open weights only2026.05 | 74.7 | — | |
| Qwen2.5-VL-32B-Instructparameters=32B, type=Instruct2025.12 | 74.59 | — | |
| Qwen2.5-VL-32B-Instructnumber of parameters=32B2026.01 | 74.45 | — | |
| InternVL3-8BAccess=Open-source general-purpose, Parameters=8B2026.03 | 73.9 | — | |
| VST-7B-SFTCategory=Open-source Spatial Intelligence Models2026.05 | 73.7 | — | |
| GR-ER 1.5Access Mode=API call only2026.05 | 73.4 | — | |
| Qwen2.5-VL-72B-Instructnumber of parameters=72B2026.01 | 73.3 | — | |
| Pelican-VL-7BAccess=Embodied Brain MLLMs, Parameters=7B2026.03 | 73.2 | — | |
| Pelican-VLParams=7B2026.06 | 73.2 | — | |
| BAGEL-7B-MoTCategory=Open-source General Models2026.05 | 73.1 | — | |
| Cambrian-S-7BScale=7B2026.05 | 72.8 | — | |
| Cambrian-S-7BCategory=Open-source Spatial Intelligence Models2026.05 | 72.8 | — | |
| Action-Sketcher-3Bnumber of parameters=3B2026.01 | 72.65 | — | |
| SenseNova-SI-1.1-Qwen3-VL-8BCategory=Open-source Spatial Intelligence Models2026.05 | 72.5 | — | |
| SR-3D (Base)Model Variant=Base2026.06 | 72.5 | — | |
| InternVL3.5-4BAccess Mode=Open weights only2026.05 | 72 | — | |
| GPT-4o-2024-11-202026.01 | 71.92 | — | |
| GPT-4oAccess=Closed-source2026.03 | 71.9 | — | |
| Qwen2.5-VL-7B-InstructCategory=Open-source General Models2026.05 | 71.8 | — | |
| ViGoRL2026.06 | 71.8 | — | |
| Lumo-1-Stage2training_stage=Stage 22025.12 | 71.68 | — | |
| Qwen2.5-VL-7B-Instructparameters=7B, type=Instruct2025.12 | 71.26 | — | |
| GenieReasonerparameters=3B, zero-shot=true2025.12 | 70.66 | — | |
| VeBrain-8Bnumber of parameters=8B2026.01 | 70.52 | — | |
| VeBrain-7BAccess=Embodied Brain MLLMs, Parameters=7B2026.03 | 70.5 | — | |
| VeBrainParams=7B2026.06 | 70.5 | — | |
| Qwen2.5-VL-7Bparameters=7B, zero-shot=true2025.12 | 70.47 | — | |
| Qwen2.5-VL-7BModel Scale=7B2026.06 | 70.4 | — | |
| VST2026.06 | 70.4 | — | |
| InternVL3.5-8BAccess=Open-source general-purpose, Parameters=8B2026.03 | 70.3 | — | |
| InternVL3.5Params=8B2026.06 | 70.3 | — | |
| LLaVA-OV-7BAccess Mode=Open weights only2026.05 | 69.7 | — | |
| SpaceR2026.06 | 69.4 | — | |
| Qwen3-VL-2B-Inst.Access=Open-source general-purpose, Parameters=2B2026.03 | 69.2 | — | |
| Cosmos-Reason1-7Bparameters=7B, zero-shot=true2025.12 | 68.9 | — | |
| VLM-3R-7BScale=7B2026.05 | 68.2 | — | |
| RoboBrain-7B-1.0number of parameters=7B, version=1.02026.01 | 68.13 | — |