Spatial Reasoning on RefSpatial
53.07Accuracy (Spatial Reasoning)SpaceTools-3B
Evaluation results
| Method | Links | |
|---|---|---|
| SpaceTools-3BModel Category=Ours2025.12 | 53.07 | |
| Molmo2Access Mode=Open weights, Open data, Open code2026.05 | 52.5 | |
| Molmo2-ERAccess Mode=Open weights, Open data, Open code2026.05 | 52.5 | |
| Qwen3-VL-4BAccess Mode=Open weights only, Model Variant=4B2026.05 | 49.5 | |
| GR-ER 1.5 ThinkingAccess Mode=API call only, Thinking=True2026.05 | 48.5 | |
| RoboRefer-8B-SFTParameters=8B, Mode=SFT2026.04 | 48.4 | |
| RoboRefer-8B-SFTModel Category=Spatial VLMs2025.12 | 48.37 | |
| Qwen3-VL-8BAccess Mode=Open weights only, Model Variant=8B2026.05 | 47.5 | |
| OpenVLThinkerV22026.04 | 44.6 | |
| Qwen3-VL-InstructMode=Instruct2026.04 | 43.9 | |
| G2VLM2026.04 | 43.5 | |
| Qwen3-VL GRPOOptimization=GRPO2026.04 | 43.3 | |
| Qwen3-VL GDPOOptimization=GDPO2026.04 | 43 | |
| GR-ER 1.5Access Mode=API call only2026.05 | 41.8 | |
| Gemini-ER 1.5Model Category=Proprietary Models2025.12 | 41.72 | |
| OneThinker-8BParameters=8B2026.04 | 38.9 | |
| Gemini 2.5 Pro2026.04 | 36.5 | |
| InternVL3.5-8BAccess Mode=Open weights only, Model Variant=8B2026.05 | 35 | |
| Gemini 2.5 ProAccess Mode=API call only, Model Variant=Pro2026.05 | 33.6 | |
| RoboBrain2.0-7BModel Category=Spatial VLMs2025.12 | 32.5 | |
| RoboBrain2.0-7BParameters=7B2026.04 | 32.5 | |
| GPT-52026.04 | 23.8 | |
| GPT-5Access Mode=API call only2026.05 | 23.5 | |
| GPT-5Model Category=Proprietary Models2025.12 | 23.1 | |
| Qwen2.5-VL-3B-Tool-freeModel Category=Tool-free Fine-tuning, Fine-tuning Protocol=RL2025.12 | 23.1 | |
| GPT-5-miniAccess Mode=API call only, Model Variant=mini2026.05 | 23 | |
| InternVL3.5-4BAccess Mode=Open weights only, Model Variant=4B2026.05 | 21.5 | |
| Qwen2.5-VL-3B-Tool-freeModel Category=Tool-free Fine-tuning, Fine-tuning Protocol=SFT2025.12 | 20.22 | |
| RoboPoint-13BModel Category=Spatial VLMs2025.12 | 15.59 | |
| GPT-4oModel Category=Proprietary Models2025.12 | 8.48 | |
| Claude Sonnet 4.5Model Category=Proprietary Models2025.12 | 7.49 | |
| Qwen2.5-VL-32BModel Category=General Open-Source Models2025.12 | 7.28 | |
| SpaceLLaVA-13BModel Category=Spatial VLMs2025.12 | 3.25 | |
| SpaceLLaVA-13BParameters=13B2026.04 | 3.25 | |
| LLaVA-OV-7BAccess Mode=Open weights only, Model Variant=7B2026.05 | 3 | |
| LLaVA-NeXT-8BModel Category=General Open-Source Models2025.12 | 0.78 | |
| Qwen2.5-VL-3BModel Category=General Open-Source Models2025.12 | 0 | |
| Molmo-7BModel Category=Spatial VLMs2025.12 | 0 |