Share your thoughts, 1 month free Claude Pro on usSee more

Multi-image Spatial Reasoning on SPAR-Bench-MV + MindCube-Tiny + MMSI-Bench (test)

51Overall Score

GPT-5.2

Updated 23d ago

Evaluation Results

Method	Links
GPT-5.2 2026.02		51
GPT-5.2 2026.07		51
OmniView-Space-R1† 2026.07		50.6
OmniView-Space-R1‡ 2026.07		46.5
Gemini-3-Pro 2026.02		44.9
Gemini-3-Pro 2026.07		44.9
HATCH 2026.02		43.6
GPT-4.1 2026.02		40.7
GPT-4.1 2026.07		40.7
OmniView-Space-Zero 2026.07		40.1
OmniView-Space-SFT 2026.07		37.9
Qwen2.5-VL-72B 2026.02		36.5
InternVL-2.5-4B 2026.02		35.5
SpatialLadder-3B 2026.02		35.4
Qwen3-VL-8B-Instruct 2026.07		35
Cambrian-S-7B 2026.07		34.9
InternVL3-8B 2026.07		34.8
Video-R1 2026.02		34.2
ViLaSR-7B 2026.07		34.1
SpatialLadder-3B 2026.07		33.9
Qwen2.5-VL-32B 2026.02		33.8
LLaVA-OneVision-4B 2026.02		33.7
Qwen3-VL-4B-Instruct 2026.07		33.1
SpaceR-7B 2026.07		33
MindCube-3B 2026.07		32.1
SpaceR-7B 2026.02		31.9
Spatial-MLLM-4B 2026.02		30.7
Qwen2.5-VL-7B 2026.02		30.4
InternVL-2.5-8B 2026.02		30.3
Qwen2.5-VL-7B-Instruct 2026.07		30.1
Spatial-MLLM-4B 2026.07		30
Qwen2.5-VL-3B 2026.02		29.4