List your company
ResearchFormerly WizWand

General-purpose Multimodal Understanding on MMStar

69.8AccuracyQwen3-VL-32B-Instruct
40.57648.16355.7563.337May 18, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen3-VL-32B-InstructBackbone model=Qwen3-VL-32B-Instruct2026.0569.8——
Vision-R1Backbone model=Qwen3-VL-4B2026.0563.8——
IVR-R1Backbone model=Qwen3-VL-4B2026.0563.3——
Vision-SR1Backbone model=Qwen3-VL-4B2026.0563.1——
IVR-R1Backbone model=Qwen2.5-VL-7B2026.0561.1——
Supervised Fine-tuning (before RL)Backbone model=Qwen3-VL-4B2026.0558.9——
Vision-R1Backbone model=Qwen2.5-VL-7B2026.0558.3——
Zero-shot Inference (before RL)Backbone model=Qwen3-VL-4B2026.0558.3——
Vision-SR1Backbone model=Qwen2.5-VL-7B2026.0557.7——
Supervised Fine-tuning (before RL)Backbone model=Qwen2.5-VL-7B2026.0556.7——
Qwen2.5-VL-32B-InstructBackbone model=Qwen2.5-VL-32B-Instruct2026.0548.3——
Qwen2.5-VL-72B-InstructBackbone model=Qwen2.5-VL-72B-Instruct2026.0545.5——
Zero-shot Inference (before RL)Backbone model=Qwen2.5-VL-7B2026.0541.7——
InternVLScale=7B2026.04—63.270.6
MUPOScale=7B2026.04—68.775.4
QwenVLScale=7B2026.04—59.274.8
R1-OVScale=7B2026.04—64.767.5
V-R1Scale=7B2026.04—66.368.9
VLAAScale=7B2026.04—66.169.1