List your company
ResearchFormerly WizWand

Multimodal Benchmarking on MMBench (Accuracy)

87.27AccuracyVISE
65.357271.046176.73582.4239Aug 25, 2025Oct 15, 2025Dec 6, 2025Jan 27, 2026Mar 20, 2026May 11, 2026Jul 2, 2026
Updated 4mo ago

Evaluation results

MethodLinks
VISEModel Scale=32B2026.0687.27
VisionZero-CLEVRModel Scale=32B2026.0687.18
VisionZero-ChartModel Scale=32B2026.0687.13
iReasonerModel Scale=32B2026.0687.08
EvoLMMModel Scale=32B2026.0687.02
VisionZero-RWModel Scale=32B2026.0686.72
BaseModel Scale=32B2026.0686.65
VisPlayModel Scale=32B2026.0686.58
VISEModel Scale=8B2026.0685.46
VisionZero-CLEVRModel Scale=8B2026.0685.11
iReasonerModel Scale=8B2026.0685.09
VisionZero-ChartModel Scale=8B2026.0685.07
EvoLMMModel Scale=8B2026.0685.02
VisionZero-CLEVRModel Scale=4B2026.0684.83
VisionZero-RWModel Scale=8B2026.0684.76
BaseModel Scale=8B2026.0684.71
VisPlayModel Scale=8B2026.0684.62
VISEModel Scale=4B2026.0684.51
InternVL2.5-8BBase Model=InternVL2.5-8B, Token Reduction Ratio=0%2026.0484.4
iReasonerModel Scale=4B2026.0684.34
EvoLMMModel Scale=4B2026.0684.21
Qwen2.5-VL-7BBase Model=Qwen2.5-VL-7B, Token Reduction Ratio=0%2026.0483.9
VisionZero-ChartModel Scale=4B2026.0683.78
VanillaBackbone=Qwen3-VL-4B-FP8, Flops Ratio Reduction=0%, Precision=FP82026.0483.6
VisPlayModel Scale=4B2026.0683.47
BaseModel Scale=4B2026.0683.42
VisionZero-RWModel Scale=4B2026.0683.32
VanillaBackbone=Qwen2-VL-7B, Flops Ratio Reduction=0%2026.0482.5
ID-SelectionBase Model=Qwen2.5-VL-7B, Token Reduction Ratio=77.8%, Importance Source=LLM second-layer cross-modal attention2026.0482.1
ID-SelectionBase Model=InternVL2.5-8B, Token Reduction Ratio=77.8%, Importance Source=LLM second-layer cross-modal attention2026.0480.9
HalfVBackbone=Qwen3-VL-4B-FP8, Flops Ratio Reduction=77.8%, Precision=FP82026.0480.9
FastVBase Model=Qwen2.5-VL-7B, Token Reduction Ratio=77.8%2026.0480.6
FastVBase Model=InternVL2.5-8B, Token Reduction Ratio=77.8%2026.0480.5
ID-SelectionBase Model=Qwen2.5-VL-7B, Token Reduction Ratio=88.9%, Importance Source=LLM second-layer cross-modal attention2026.0480
DARTBase Model=InternVL2.5-8B, Token Reduction Ratio=77.8%2026.0480
DARTBase Model=Qwen2.5-VL-7B, Token Reduction Ratio=77.8%2026.0479.8
MM1LLM Backbone=MM1-7B-MoE2026.0579.7
HalfVBackbone=Qwen2-VL-7B, Flops Ratio Reduction=88.9%2026.0479.2
DARTBackbone=Qwen3-VL-4B-FP8, Flops Ratio Reduction=77.8%, Precision=FP82026.0479.2
ID-SelectionBase Model=InternVL2.5-8B, Token Reduction Ratio=88.9%, Importance Source=LLM second-layer cross-modal attention2026.0478.8
FastVBackbone=Qwen3-VL-4B-FP8, Flops Ratio Reduction=77.8%, Precision=FP82026.0477.5
HalfVBackbone=Qwen3-VL-4B-FP8, Flops Ratio Reduction=88.9%, Precision=FP82026.0477.4
AsyMoE-LLaMA3LLM Backbone=LLaMA-3-8B, Act Param=10.8B2026.0576.8
VISEModel Scale=2B2026.0676.72
DARTBackbone=Qwen3-VL-4B-FP8, Flops Ratio Reduction=88.9%, Precision=FP82026.0476.2
AsyMoE-Phi3LLM Backbone=Phi-3-mini, Act Param=4.1B2026.0575.9
FastVBase Model=InternVL2.5-8B, Token Reduction Ratio=88.9%2026.0475.8
MoIIE-LLaMA3LLM Backbone=LLaMA-3-8B, Act Param=11.3B2026.0575.7
Mini-GeminiLLM Backbone=Mixtral-8×7B, Act Param=13.5B2026.0575.6
MoIIE-Phi3LLM Backbone=Phi-3-mini, Act Param=5.5B2026.0575.4
CuMoLLM Backbone=Mixtral-8×7B, Act Param=13.5B2026.0575.3
VisionZero-CLEVRModel Scale=2B2026.0675.07
DARTBase Model=InternVL2.5-8B, Token Reduction Ratio=88.9%2026.0475
FastVBackbone=Qwen3-VL-4B-FP8, Flops Ratio Reduction=88.9%, Precision=FP82026.0474.9
ContextMoE-Phi2LLM Backbone=Phi-2, Act Param=4.1B2026.0574.9
DARTBackbone=Qwen2-VL-7B, Flops Ratio Reduction=88.9%2026.0474.8
iReasonerModel Scale=2B2026.0674.75
EvoLMMModel Scale=2B2026.0674.62
VisPlayModel Scale=2B2026.0674.52
VisionZero-ChartModel Scale=2B2026.0674.51
BaseModel Scale=2B2026.0674.48
VisionZero-RWModel Scale=2B2026.0674.4
DARTBase Model=Qwen2.5-VL-7B, Token Reduction Ratio=88.9%2026.0474.1
FastVBase Model=Qwen2.5-VL-7B, Token Reduction Ratio=88.9%2026.0473.8
CapRL-Image-5MModel Configuration=InternLM2.5-7B + CLIP-ViT-L2026.0673.7
CapRL-Image-1MModel Configuration=InternLM2.5-7B + CLIP-ViT-L2026.0673.4
Deepseek-VLLLM Backbone=Deepseek-7B2026.0573.2
CapRL-Image-5MModel Configuration=Qwen2.5-3B + Qwen2.5-ViT2026.0673.1
CapRL-Image-5MModel Configuration=Qwen2.5-7B + Qwen2.5-ViT2026.0672.7
DenseFusion-1MModel Configuration=Qwen2.5-7B + Qwen2.5-ViT2026.0672.6
ShareGPT4V-1MModel Configuration=InternLM2.5-7B + CLIP-ViT-L2026.0672.6
HoloVBackbone=Qwen2-VL-7B, Flops Ratio Reduction=88.9%2026.0472.4
LLaVA-LLaMA3LLM Backbone=LLaMA-3-8B, Act Param=8.4B2026.0572.3
DenseFusion-1MModel Configuration=InternLM2.5-7B + CLIP-ViT-L2026.0672.2
VanillaModel Configuration=Qwen2.5-7B + Qwen2.5-ViT2026.0672.1
CapRL-Image-1MModel Configuration=Qwen2.5-7B + Qwen2.5-ViT2026.0672.1
ShareGPT4V-1MModel Configuration=Qwen2.5-7B + Qwen2.5-ViT2026.0671.9
SPHINX-XLLM Backbone=Mixtral-8×7B2026.0571.3
VanillaModel Configuration=InternLM2.5-7B + CLIP-ViT-L2026.0670.7
CapRL-Image-1MModel Configuration=Qwen2.5-3B + Qwen2.5-ViT2026.0670.5
LLaVA-NeXT Vanilla 13BToken Budget=2880, Avg. Images=4.12, Avg. Tokens=2373.122025.0870
Mini-GeminiLLM Backbone=Vicuna-7B, Act Param=7.3B2026.0569.3
FastVBackbone=Qwen2-VL-7B, Flops Ratio Reduction=88.9%2026.0469.2
DenseFusion-1MModel Configuration=Qwen2.5-3B + Qwen2.5-ViT2026.0669
ShareGPT4V-1MModel Configuration=Qwen2.5-3B + Qwen2.5-ViT2026.0668.9
ShareGPT-4VLLM Backbone=Vicuna-7B2026.0568.8
MousiLLM Backbone=Vicuna-7B, Act Param=7.9B2026.0568.8
VisionZipToken Budget=640, Avg. Tokens=5272025.0868.6
VanillaModel Configuration=Qwen2.5-3B + Qwen2.5-ViT2026.0668.6
MoE-LLaVA-Phi2LLM Backbone=Phi-2, Act Param=3.6B2026.0568
MMTokToken Budget=640, Avg. Tokens=5272025.0867.44
LLaVA-NeXT-7BBase Model=LLaVA-NeXT-7B, Retained Tokens=2880, Pruning Ratio=100%2026.0467.4
LLaVA-NeXTLLM Backbone=Vicuna-7B, Act Param=7.1B2026.0567.4
VisionZipToken Budget=320, Avg. Tokens=2642025.0867.2
HiPruneRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.0767
VisionZip ✨Token Budget=320, Avg. Tokens=2642025.0866.9
DivPruneToken Budget=640, Avg. Tokens=5272025.0866.84
VisionZip ✨Token Budget=640, Avg. Tokens=5272025.0866.6
EADPRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.0766.5
CDPrunerRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.0766.2