List your company
ResearchFormerly WizWand

Spatial Understanding on MMSI-Bench (Accuracy)

45.9AccuracyGemini-3-Pro
19.48426.34233.240.058Apr 8, 2026Apr 16, 2026Apr 24, 2026May 2, 2026May 10, 2026May 18, 2026May 27, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Gemini-3-Pro2026.0545.9
SenseNova-SI-8BParameters=8B2026.0543.3
GEM-8BParameters=8B2026.0535.3
Seed1.82026.0534.6
HY-EmbodiedModel Version=0.5 MoT-2B, Thinking Mode=thinking2026.0433.2
Qwen3-VL-8B-SFTParameters=8B, Training=SFT2026.0532.8
VST-7BParameters=7B2026.0532.6
MiMo-EmbodiedNumber of Parameters=7B2026.0431.9
GEM-2BParameters=2B2026.0530.6
GPT-4o2026.0530.3
Qwen3-VL-2B-SFTParameters=2B, Training=SFT2026.0528.9
InternVL3.5-8BParameters=8B2026.0528.4
VLM-3R2026.0527.9
Qwen3-VL-8BParameters=8B2026.0527.7
LLava-OneVision-7BParameters=7B2026.0526.6
SpaceR-7BParameters=7B2026.0526.4
QwenModel Version=3-VL, Number of Parameters=4B2026.0425.1
CambrainS-7BParameters=7B2026.0524.2
QwenModel Version=3-VL, Number of Parameters=2B2026.0423.6
Qwen3-VL-2BParameters=2B2026.0523.6
RoboBrainModel Version=2.5, Number of Parameters=4B2026.0420.5