List your company
ResearchFormerly WizWand

Visual Understanding on MME

2,321MME ScoreQwen2-VL
-63.408555.6211,174.651,793.679Dec 1, 2024Feb 18, 2025May 8, 2025Jul 26, 2025Oct 13, 2025Dec 31, 2025Mar 20, 2026
Updated 7mo ago

Evaluation results

MethodLinks
Qwen2-VLModel Size=8B2025.012,321
InternVL2.5-MPOModel Size=8B2025.012,321
SAIL-VLModel Size=8B2025.012,244
DeepSeekVL-2Model Size=8B2025.012,149
InternVL2.5-MPOModel Size=2B2025.012,123
SAIL-VLModel Size=2B2025.011,969
ShareGPT4VLLM=Vicuna-7B, Resolution=3362026.031,943.8
Qwen2-VLModel Size=2B2025.011,923
DeepSeekVL-2Model Size=2B2025.011,910
EvoTokLLM=Qwen-2.5-7B-Inst., Resolution=2562026.031,895.1
InternVL3.5-1B# Total Params=1.1B2025.111,894
Qwen-VL-ChatLLM=Qwen-7B, Resolution=4482026.031,848.3
LLaVA-v1.5LLM=Vicuna-1.5-13B, Resolution=3362026.031,826.7
LFM2-VL-1.6B# Total Params=1.6B2025.111,757
Emu2-ChatType=Understanding-only, LLM Params=33B2025.031,678
TokenFlow-BLLM=Vicuna-13B, Resolution=2242026.031,660.4
TokenFlow-LLLM=Vicuna-13B, Resolution=2562026.031,622.9
Dynamic-LLaVA-13B^ITFree=false, Image (prefill) Token=115, TFLOPs=4.72024.121,563.3
Dynamic-LLaVA-13B^VFree=false, Image (prefill) Token=115, TFLOPs=4.72024.121,554.1
VILALLM=LLaMA-2-7B, Resolution=3362026.031,533
LLaVA-1.5-13BImage (prefill) Token=576, TFLOPs=19.62024.121,531.3
LLaVA-1.5-7BImage (prefill) Token=576, TFLOPs=10.12024.121,510.7
DualTokenLLM=LLaMA-2-7B, Resolution=2562026.031,502.7
Dynamic-LLaVA-7B^ITFree=false, Image (prefill) Token=115, TFLOPs=2.52024.121,501
QLIPLLM=Vicuna-1.5-7B, Resolution=3922026.031,498.3
LLaVA-PruMerge+ (13B)Image (prefill) Token=146, TFLOPs=4.92024.121,485.5
Dynamic-LLaVA-7B^VFree=false, Image (prefill) Token=115, TFLOPs=2.52024.121,479.8
LLaVA-FastV (13B)Image (prefill) Token=144, TFLOPs=62024.121,470.3
LLaVA-PruMerge+ (7B)Image (prefill) Token=146, TFLOPs=2.52024.121,462.4
LLaVA-FastV (7B)Free=true, Image (prefill) Token=144, TFLOPs=3.22024.121,458.9
LLaVA-1.5-7B+H2Ok-10, -0.5Image (prefill) Token=576, TFLOPs=10.12024.121,458.4
SmolVLM2-500M# Total Params=0.5B2025.111,456
LLaVA-1.5-13b+H2Ok-10,-0.5Image (prefill) Token=576, TFLOPs=19.62024.121,448.3
UniTokLLM=LLaMA-2-7B, Resolution=2562026.031,448
VILA-ULLM=LLaMA-7B, Resolution=3842026.031,401.8
JanusLLM=DeepSeek-1.3B, Resolution=3842026.031,338
VILA-ULLM=LLaMA-7B, Resolution=2562026.031,336.2
LLaVA-FastV† (7B)Free=false, Image (prefill) Token=144, TFLOPs=3.22024.121,292.2
LFM2-VL-450M# Total Params=0.45B2025.111,230
SeGroSFine-Tuning=Ours2026.031,217
SFTFine-Tuning=SFT2026.031,210
Base ModelFine-Tuning=w/o SFT2026.031,195
RecaFine-Tuning=Reca2026.031,190
IdeficsType=Understanding-only, LLM Params=9B2025.031,177.3
MiniGPT4Type=Understanding-only, LLM Params=7B2025.031,047.4
Kosmos-2Type=Understanding-only, LLM Params=2B2025.03721.1
LLaVA-1.5-7B+H2O-0.5Image (prefill) Token=576, TFLOPs=11.62024.12500.5
LaViDa-O2025.12488
Qwen-VLType=Understanding-only, LLM Params=7B2025.03482.7
Sparse-LaViDa2025.12450
WeGenType=Unified (Understanding & Generation), LLM Params=7B2025.03447.4
Chameleon-7bType=Unified (Understanding & Generation), LLM Params=7B2025.03202.7
Kosmos-GType=Unified (Understanding & Generation), LLM Params=1.9B2025.03104.3
LLaVAType=Understanding-only, LLM Params=7B2025.0328.3