List your company

Multimodal Reasoning on MMMU (Accuracy)

72.9AccuracyGPT-4o
29.11640.48351.8563.217Jun 11, 2025Aug 14, 2025Oct 17, 2025Dec 20, 2025Feb 22, 2026Apr 27, 2026Jun 30, 2026

Evaluation results

MethodLinks
GPT-4oModel Category=Proprietary Models2025.0672.9
Gemini-2.0-ProModel Category=Proprietary Models2025.0672.6
Claude-3.7-SonnetModel Category=Proprietary Models2025.0671
InternVL2.5-78BModel Category=Open-source Models (>70B)2025.0670.1
InternVL2.5-78B-MPOModel Category=Open-source Models (>70B)2025.0668.2
Qwen2.5-VL-72BModel Category=Open-source Models (>70B)2025.0668.2
PRCRModel=InternVL3.5, #Params=8B2026.0666.14
Token-ReplayModel=InternVL3.5, #Params=8B2026.0665.82
BaselineModel=InternVL3.5, #Params=8B2026.0665.05
PRCRModel=Qwen3-VL, #Params=32B2026.0664.89
Token-ReplayModel=Qwen3-VL, #Params=32B2026.0664.73
BaselineModel=Qwen3-VL, #Params=32B2026.0663.68
PRCRModel=InternVL3.5, #Params=14B2026.0663.2
Token-ReplayModel=InternVL3.5, #Params=14B2026.0662.88
BaselineModel=InternVL3.5, #Params=14B2026.0662.02
SVIPPRM Trainable Params=7B2026.0661.3
Athena-7BModel Category=Open-source Models (~7B)2025.0661.1
Visual-PRMPRM Trainable Params=7B2026.0660.2
PRCRModel=Qwen3-VL, #Params=8B2026.0660.07
Token-ReplayModel=Qwen3-VL, #Params=8B2026.0659.45
STRIDEBase model=Qwen2.5-VL-7B2026.0658.7
Qwen2.5-VL-7BModel Category=Open-source Models (~7B)2025.0658
SRPOBase model=Qwen2.5-VL-7B2026.0657.1
VL-PRM300KPRM Trainable Params=8B2026.0656.5
InternVL2.5-8BModel Category=Open-source Models (~7B)2025.0656.2
BaselineModel=Qwen3-VL, #Params=8B2026.0655.22
MM-EurekaBase model=Qwen2.5-VL-7B2026.0655.2
MMS-PRMPRM Trainable Params=0M2026.0654.2
BaseBase model=Qwen2.5-VL-7B2026.0654.1
Op-SkipBackbone=Qwen3-VL-8B, TFLOPs=66%, N=202026.0652.89
APETBackbone=Qwen3-VL-8B, TFLOPs=66%2026.0652.53
V2DropBackbone=Qwen3-VL-8B, TFLOPs=66%2026.0652.44
ShortVBackbone=Qwen3-VL-8B, TFLOPs=58%, N=202026.0652.22
AdaSteerModel=QWEN3.52026.0652
MARSModel=INTERNVL 3.52026.0651.4
VanillaBackbone=Qwen3-VL-8B, TFLOPs=100%2026.0651.33
VSkipBackbone=Qwen3-VL-8B, TFLOPs=76%, N=202026.0651.33
Zero-shotModel=INTERNVL 3.52026.0651.3
ECSOModel=INTERNVL 3.52026.0651.3
MiniCPM-o-2.6-8BModel Category=Open-source Models (~7B)2025.0650.9
AdaSteerModel=INTERNVL 3.52026.0650.9
Zero-shotModel=QWEN3-VL2026.0650.8
ECSOModel=QWEN3-VL2026.0650.8
MARSModel=QWEN3-VL2026.0650.8
Zero-shotModel=QWEN3.52026.0650.8
ECSOModel=QWEN3.52026.0650.8
AdaSteerModel=QWEN3-VL2026.0650.4
AdaSteerModel=MOLMO22026.0650.2
APETBackbone=Qwen2.5-VL-7B, TFLOPs=55%2026.0650.11
Zero-shotModel=MOLMO22026.0650.1
ECSOModel=MOLMO22026.0650.1
MARSModel=MOLMO22026.0650.1
VanillaBackbone=Qwen2.5-VL-7B, TFLOPs=100%2026.0650
V2DropBackbone=Qwen2.5-VL-7B, TFLOPs=55%2026.0649.89
SASAModel=QWEN3-VL2026.0649.2
SASAModel=MOLMO22026.0648.2
VTWBackbone=Qwen3-VL-8B, TFLOPs=58%, K=162026.0647.33
VSkipBackbone=Qwen2.5-VL-7B, TFLOPs=67%, N=162026.0647
Op-SkipBackbone=Qwen2.5-VL-7B, TFLOPs=55%, N=202026.0646.89
VTWBackbone=Qwen2.5-VL-7B, TFLOPs=50%, K=162026.0643.56
ShortVBackbone=Qwen2.5-VL-7B, TFLOPs=50%, N=202026.0641.89
SASAModel=INTERNVL 3.52026.0641.3
Zero-shotModel=GEMMA32026.0639.8
ECSOModel=GEMMA32026.0639.8
AdaSteerModel=GEMMA32026.0639.8
MARSModel=GEMMA32026.0639.8
SASAModel=GEMMA32026.0639.4
MARSModel=QWEN3.52026.0638.2
VSkip+Backbone=LLaVA-1.5-7B, TFLOPs=76%, N=202026.0636.44
VanillaBackbone=LLaVA-1.5-7B, TFLOPs=100%2026.0636.11
APETBackbone=LLaVA-1.5-7B, TFLOPs=66%2026.0636.11
VTWBackbone=LLaVA-1.5-7B, TFLOPs=55%, K=162026.0635.67
ShortVBackbone=LLaVA-1.5-7B, TFLOPs=55%, N=202026.0635.67
Op-SkipBackbone=LLaVA-1.5-7B, TFLOPs=66%, N=202026.0635.67
VSkipBackbone=LLaVA-1.5-7B, TFLOPs=76%, N=202026.0635.56
V2DropBackbone=LLaVA-1.5-7B, TFLOPs=66%2026.0631.67
SASAModel=QWEN3.52026.0630.8