List your company

Multi-discipline reasoning on MMMU (val) (Accuracy)

81.8AccuracyGPT-5
28.44842.29956.1570.001Feb 12, 2026Mar 1, 2026Mar 19, 2026Apr 6, 2026Apr 23, 2026May 11, 2026May 29, 2026

Evaluation results

MethodLinks
GPT-5Access=API call only2026.0381.8
Gemini 2.5 ProAccess=API call only2026.0379.6
Gemini 2.5 FlashAccess=API call only2026.0379
GPT-5 miniAccess=API call only2026.0378.7
Claude Sonnet 4.5Access=API call only2026.0377.8
InternVL3.5-8BAccess=Open weights only2026.0373.4
Keye-VL-1.5-8BAccess=Open weights only2026.0371.4
Qwen3-VL-8BAccess=Open weights only2026.0369.6
GLM-4.1V-9BAccess=Open weights only2026.0368
MiniCPM-V-4.5-8BAccess=Open weights only2026.0367.7
Qwen3-VL-32BBackbone=Qwen3-VL, Model Size=32B, Training Method=Baseline2026.0567.7
iVGR-Qwen3-VL-32BBackbone=Qwen3-VL, Model Size=32B, Training Method=iVGR2026.0567.7
Qwen3-VL-4BAccess=Open weights only2026.0367.4
InternVL3.5-4BAccess=Open weights only2026.0366.6
iVGR-Qwen3-VL-8BBackbone=Qwen3-VL, Model Size=8B, Training Method=iVGR2026.0559.8
Qwen3-VL-8BBackbone=Qwen3-VL, Model Size=8B, Training Method=Baseline2026.0558
Qwen3-VL-8BMode=base, Extractor=Regex letter extractor2026.0556.8
S^3-FTBackbone=Qwen3-VL-8B, Mode=Single-view, Extractor=Regex letter extractor2026.0556
Eagle2.5-8BAccess=Open weights only2026.0355.8
iVGR-Qwen2.5-VL-7BBackbone=Qwen2.5-VL, Model Size=7B, Training Method=iVGR2026.0555.2
S^3-FTBackbone=Qwen3-VL-8B, Mode=Cross-viewpoint, Extractor=Regex letter extractor2026.0554.9
Qwen2.5-VL-7BBackbone=Qwen2.5-VL, Model Size=7B, Training Method=Baseline2026.0554.4
MolmoPoint-8BAccess=MolmoPoint2026.0353.7
Molmo2-8BAccess=Fully open2026.0353
Qwen2.5-VL-7BMode=base, Extractor=Regex letter extractor2026.0552.4
S^3-FTBackbone=Qwen2.5-VL-7B, Mode=Single-view, Extractor=Regex letter extractor2026.0551
Molmo2-4BAccess=Fully open2026.0350.9
S^3-FTBackbone=Qwen2.5-VL-7B, Mode=Cross-viewpoint, Extractor=Regex letter extractor2026.0549.6
PLM-8BAccess=Fully open2026.0346.1
MolmoPoint-8B-O-7BAccess=Fully open2026.0345.8
PLM-3BAccess=Fully open2026.0341.2
Qwen2.5-VL-72B InstructZero-shot=true2026.0237.2
Qwen2.5-VL-32B + AT-RL (Ours)Zero-shot=true2026.0236.5
Qwen2.5-VL-32B + VPPOZero-shot=true2026.0234.1
Gemini 2.0 FlashZero-shot=true2026.0232.9
Qwen2.5-VL-32B InstructZero-shot=true2026.0231.7
OpenAI GPT-4oZero-shot=true2026.0231.1
Claude 3.5 SonnetZero-shot=true2026.0230.5