List your company
ResearchFormerly WizWand

Multi-modal Understanding on MMBench EN

86.3Overall ScoreGemini-2.5-Pro
30.24444.79759.3573.903Aug 9, 2024Nov 8, 2024Feb 8, 2025May 11, 2025Aug 10, 2025Nov 10, 2025Feb 10, 2026
Updated 7mo ago

Evaluation results

MethodLinks
Gemini-2.5-ProDate=2024.12, Tokenization Paradigm=Proprietary Models2026.0286.3
VQ-RAEModel Scale=7B, Date=2025.11, Tokenization Paradigm=Hybrid Tokenization2026.0285.1
BagelModel Scale=14B, Date=2025.05, Tokenization Paradigm=Continuous Tokenization2026.0285
Bagel#LLM=14B2025.1285
STAR-7B#LLM=7B2025.1283.9
BLIP3-o#LLM=8B2025.1283.5
MetaQuery-XL#LLM=7B2025.1283.5
VanillaMode=Upper bound2024.1283.5
ManzanoModel Scale=30B, Date=2025.09, Tokenization Paradigm=Hybrid Tokenization2026.0283.4
iLLaVAImage token reduction ratio=66.7%2024.1283
iLLaVAImage token reduction ratio=77.8%2024.1282.9
SparseVLMImage token reduction ratio=66.7%2024.1282.7
Qwen2.5VLModel Scale=7B, Date=2025.02, Tokenization Paradigm=Understanding-only Models2026.0282.6
VisionZipImage token reduction ratio=66.7%2024.1282.5
PyramidDropImage token reduction ratio=66.7%2024.1281.6
FasterVLMImage token reduction ratio=66.7%2024.1281.2
ILLUME+#LLM=3B2025.1280.8
SparseVLMImage token reduction ratio=77.8%2024.1280.3
KelixModel Scale=8B, Tokenization Paradigm=Discrete Tokenization2026.0280.2
STAR-3B#LLM=3B2025.1280.1
iLLaVAImage token reduction ratio=88.9%2024.1279.7
FasterVLMImage token reduction ratio=77.8%2024.1279.6
VisionZipImage token reduction ratio=77.8%2024.1279.5
Show-o2#LLM=7B2025.1279.3
Janus-ProModel Scale=7B, Date=2025.01, Tokenization Paradigm=Discrete Tokenization2026.0279.2
Janus-Pro#LLM=7B2025.1279.2
VisionZipImage token reduction ratio=88.9%2024.1279.2
FasterVLMImage token reduction ratio=88.9%2024.1278.7
Blip-3oModel Scale=4B, Date=2025.05, Tokenization Paradigm=Continuous Tokenization2026.0278.6
PyramidDropImage token reduction ratio=77.8%2024.1278.5
SparseVLMImage token reduction ratio=88.9%2024.1278.2
Ovis-U1#LLM=1.5B2025.1277.8
MiniCPM-Llama3-V2.5Number of Parameters=8B, Evaluation Mode=Zero-shot2024.0877.6
mPLUG-Owl3Number of Parameters=8B, Evaluation Mode=Zero-shot2024.0877.6
EVLM-ChatNumber of Parameters=32B, Evaluation Mode=Zero-shot2024.0876.9
Qwen2.5VLModel Scale=3B, Date=2025.02, Tokenization Paradigm=Understanding-only Models2026.0276.4
Idefics2Number of Parameters=8B, Evaluation Mode=Zero-shot2024.0875.7
Janus-ProModel Scale=1.5B, Date=2025.01, Tokenization Paradigm=Discrete Tokenization2026.0275.5
Janus#LLM=1.3B2025.1275.5
MetaMorph#LLM=8B2025.1275.2
X-OmniModel Scale=7B, Date=2025.07, Tokenization Paradigm=Discrete Tokenization2026.0274.8
X-Omni#LLM=7B2025.1274.8
CambrianNumber of Parameters=8B, Evaluation Mode=Zero-shot2024.0874.6
PyramidDropImage token reduction ratio=88.9%2024.1274.4
VILA1.5Number of Parameters=8B, Evaluation Mode=Zero-shot2024.0872.3
Seed-X#LLM=13B2025.1270.1
Mantis-SigLIPNumber of Parameters=8B, Evaluation Mode=Zero-shot2024.0868.7
LLAVA-NextNumber of Parameters=8B, Evaluation Mode=Zero-shot2024.0867.4
CogVLMNumber of Parameters=17B, Evaluation Mode=Zero-shot2024.0865.8
mPLUG-Owl2Number of Parameters=8B, Evaluation Mode=Zero-shot2024.0864.5
LLAVA-1.5Number of Parameters=8B, Evaluation Mode=Zero-shot2024.0864.3
Emu3Model Scale=8B, Date=2024.09, Tokenization Paradigm=Discrete Tokenization2026.0258.5
EMU3#LLM=8B2025.1258.5
InstructBLIPNumber of Parameters=13B, Evaluation Mode=Zero-shot2024.0838.3
OpenFlamingoNumber of Parameters=9B, Evaluation Mode=Zero-shot2024.0832.4