List your company
ResearchFormerly WizWand

Expert-level Multimodal Understanding on MMMU (Accuracy)

60.9AccuracyQwen3-VL-8B
42.49247.27152.0556.829Sep 30, 2025Oct 8, 2025Oct 16, 2025Oct 24, 2025Nov 1, 2025Nov 9, 2025Nov 17, 2025
Updated 7mo ago

Evaluation results

MethodLinks
Qwen3-VL-8BModel Category=Open-source General Models2025.1160.9
PAPOModel Scale=7B, Decoding Strategy=greedy decoding2025.0960.8
VAPOModel Scale=7B2025.0960.2
VL-RethinkerModel Scale=7B2025.0957.9
SenseNova-SIBase Architecture=Qwen3-VL-8B, Model Category=Ours2025.1157.6
MM-EurekaModel Scale=7B2025.0955.7
InternVL3-8BModel Category=Open-source General Models2025.1155.6
Base ModelModel Scale=7B2025.0952.7
VST-7B-SFTModel Category=Open-source SI Models2025.1150.9
Bagel-7B-MoTModel Category=Open-source General Models2025.1150.4
SenseNova-SIBase Architecture=Bagel-7B-MoT, Model Category=Ours2025.1150.2
SenseNova-SIBase Architecture=InternVL3-8B, Model Category=Ours2025.1149.4
Cambrian-S-7BModel Category=Open-source SI Models2025.1147.1
SenseNova-SIBase Architecture=InternVL3-2B, Model Category=Ours2025.1144.4
InternVL3-2BModel Category=Open-source General Models2025.1143.2