List your company
ResearchFormerly WizWand

Spatial Reasoning (Multi-Image) on ERQA

51.02AccuracyGemini2.5-Pro
11.999222.129632.2642.3904May 22, 2025Jul 5, 2025Aug 18, 2025Oct 1, 2025Nov 14, 2025Dec 28, 2025Feb 11, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Gemini2.5-ProParams=–2025.1151.02
MiMo-EmbodiedParams=7B2025.1146.75
Claude-Sonnet-4Params=–2025.1145.75
ViLAVT-7BThinking Level=Chatting with Images2026.0242.2
MiMo-VLParams=7B2025.1141.5
VILASR-7BThinking Level=Thinking with Images2026.0241
InternVL3.5Params=8B2025.1141
Qwen-VL-MaxParams=–2025.1140.91
SpaceR-7BThinking Level=Thinking about Images2026.0240.3
Qwen2.5-VL-7BThinking Level=Thinking about Images2026.0239.8
Qwen2.5-VL-7BThinking Level=Non-Thinking2026.0239.3
Cosmos-Reason1Params=7B2025.1139.09
Qwen2.5-VLParams=7B2025.1138.8
Spatial-MLLM-4BThinking Level=Thinking about Images2026.0238.3
VeBrainParams=8B2025.1137.29
RoboBrain-1.0Params=7B2025.1136.52
InternVL3-8BThinking Level=Non-Thinking2026.0235.3
GPT-4oParams=–2025.1132.48
LLaVA-OneVision-7BThinking Level=Non-Thinking2026.0230.6
RoboBrain-2.0Params=7B2025.1130.31
MagmaParams=8B2025.1125.73
Multi-SpatialMLLMModel=Multi-SpatialMLLM2025.0521.6
InternVL-8BModel=InternVL-8B2025.0513.5