Multimodal Understanding on MMStar (Score)
68.33ScoreSGT-BAGEL
Evaluation Results
| Method | Links | |
|---|---|---|
| SGT-BAGELParams=7B+7B‡2026.05 | 68.33 | |
| BAGELParams=7B+7B‡2026.05 | 67.46 | |
| Qwen2.5-VL-VIFParams=7B, Base model=Qwen-2.5-VL2026.05 | 65.2 | |
| UniWorld-v1Params=7B+12B‡2026.05 | 63.9 | |
| VanillaVisual Tokens=360, Prefill Time (s)=0.1337, FLOPs (T)=0.9802, GPU Mem. (GB)=15.97, SpeedUp=1.00×2026.05 | 62.9 | |
| Qwen2.5-VL-SFTParams=7B, Base model=Qwen-2.5-VL2026.05 | 62.53 | |
| Qwen2.5-VLParams=7B, Base model=Qwen-2.5-VL2026.05 | 62.06 | |
| Gemini ProParams=-2026.05 | 59.1 | |
| mPLUG-Owl3Params=8B2026.05 | 59.1 | |
| SGT-Gen2Params=3B+4B‡2026.05 | 57.07 | |
| NGMModel=Qwen3-VL-2B2026.05 | 56.2 | |
| OmniGen2Params=3B+4B‡2026.05 | 55.07 | |
| Qwen3-VL-2BModel=Qwen3-VL-2B2026.05 | 54.67 | |
| SPprunerVisual Tokens=80, Prefill Time (s)=0.0618, FLOPs (T)=0.3688, GPU Mem. (GB)=15.74, SpeedUp=2.16×2026.05 | 50.7 | |
| DARTVisual Tokens=80, Prefill Time (s)=0.0833, FLOPs (T)=0.3688, GPU Mem. (GB)=15.74, SpeedUp=1.60×2026.05 | 49.7 | |
| Janus-ProParams=7B2026.05 | 46.8 | |
| LLaVA-1.5-VIFParams=13B, Base model=LLaVA-1.52026.05 | 41.07 | |
| LLaVA-1.5-SFTParams=13B, Base model=LLaVA-1.52026.05 | 39.93 | |
| HarmonParams=1.5B2026.05 | 38 | |
| LLaVA-1.5-VIFParams=7B, Base model=LLaVA-1.52026.05 | 37.07 | |
| VISTAParams=7B, Base model=LLaVA-1.52026.05 | 35.91 | |
| LLaVA-1.5-SFTParams=7B, Base model=LLaVA-1.52026.05 | 35.53 | |
| LLaVA-1.5Params=13B, Base model=LLaVA-1.52026.05 | 33.87 | |
| LLaVA-1.5Params=7B, Base model=LLaVA-1.52026.05 | 33 | |
| ChameleonParams=7B2026.05 | 28.93 | |
| ReCA-HarmonParams=1.5B2026.05 | 25.53 |