Multimodal Understanding on MMMU, SEED, OCRBench, VizWiz, ScienceQA, and TextVQA (test)
78.1Average AccuracyFP16
Evaluation Results
| Method | Links | |
|---|---|---|
| FP16Bit=FP16, Model Architecture=Qwen2-VL-72B2026.05 | 78.1 | |
| MGVQBit=3, Model Architecture=Qwen2-VL-72B2026.05 | 77.8 | |
| MBQBit=3, Model Architecture=Qwen2-VL-72B2026.05 | 77.6 | |
| AWQBit=3, Model Architecture=Qwen2-VL-72B2026.05 | 77.5 | |
| VPTQBit=3, Model Architecture=Qwen2-VL-72B2026.05 | 77.5 | |
| GPTQBit=3, Model Architecture=Qwen2-VL-72B2026.05 | 76.6 | |
| MGVQBit=2, Model Architecture=Qwen2-VL-72B2026.05 | 75.8 | |
| RTNBit=3, Model Architecture=Qwen2-VL-72B2026.05 | 75 | |
| FP16Bit=FP16, Model Architecture=InternVL2-26B2026.05 | 74.6 | |
| FP16Bit=FP16, Model Architecture=LLaVA-onevision-72B2026.05 | 74.3 | |
| MGVQBit=3, Model Architecture=InternVL2-26B2026.05 | 74.2 | |
| MGVQBit=3, Model Architecture=LLaVA-onevision-72B2026.05 | 74 | |
| VPTQBit=3, Model Architecture=InternVL2-26B2026.05 | 73.9 | |
| QuIP#Bit=2, Model Architecture=Qwen2-VL-72B2026.05 | 73.9 | |
| MBQBit=3, Model Architecture=InternVL2-26B2026.05 | 73.8 | |
| MBQBit=3, Model Architecture=LLaVA-onevision-72B2026.05 | 73.6 | |
| VPTQBit=3, Model Architecture=LLaVA-onevision-72B2026.05 | 73.6 | |
| AWQBit=3, Model Architecture=InternVL2-26B2026.05 | 73.5 | |
| RTNBit=3, Model Architecture=InternVL2-26B2026.05 | 73.3 | |
| FP16Bit=FP16, Model Architecture=Qwen2-VL-7B2026.05 | 73.1 | |
| VPTQBit=2, Model Architecture=Qwen2-VL-72B2026.05 | 72.6 | |
| MGVQBit=2, Model Architecture=LLaVA-onevision-72B2026.05 | 72.4 | |
| GPTQBit=3, Model Architecture=LLaVA-onevision-72B2026.05 | 72.3 | |
| GPTQBit=3, Model Architecture=InternVL2-26B2026.05 | 72.3 | |
| RTNBit=3, Model Architecture=LLaVA-onevision-72B2026.05 | 72.1 | |
| MGVQBit=3, Model Architecture=Qwen2-VL-7B2026.05 | 71.9 | |
| FP16Bit=FP16, Model Architecture=InternVL2-8B2026.05 | 71.7 | |
| MGVQBit=2, Model Architecture=InternVL2-26B2026.05 | 71.4 | |
| VPTQBit=3, Model Architecture=Qwen2-VL-7B2026.05 | 71.3 | |
| MGVQBit=3, Model Architecture=InternVL2-8B2026.05 | 71.1 | |
| MBQBit=2, Model Architecture=Qwen2-VL-72B2026.05 | 71.1 | |
| MBQBit=3, Model Architecture=Qwen2-VL-7B2026.05 | 70.9 | |
| QuIP#Bit=2, Model Architecture=LLaVA-onevision-72B2026.05 | 70.9 | |
| VPTQBit=3, Model Architecture=InternVL2-8B2026.05 | 70.6 | |
| MBQBit=3, Model Architecture=InternVL2-8B2026.05 | 70.4 | |
| AWQBit=3, Model Architecture=Qwen2-VL-7B2026.05 | 70.3 | |
| VPTQBit=2, Model Architecture=LLaVA-onevision-72B2026.05 | 70.2 | |
| MGVQBit=2, Model Architecture=Qwen2-VL-7B2026.05 | 70 | |
| AWQBit=3, Model Architecture=InternVL2-8B2026.05 | 69.8 | |
| RTNBit=3, Model Architecture=InternVL2-8B2026.05 | 69 | |
| QuIP#Bit=2, Model Architecture=Qwen2-VL-7B2026.05 | 68.2 | |
| GPTQBit=3, Model Architecture=Qwen2-VL-7B2026.05 | 67.9 | |
| VPTQBit=2, Model Architecture=Qwen2-VL-7B2026.05 | 67.3 | |
| GPTQBit=3, Model Architecture=InternVL2-8B2026.05 | 67.2 | |
| GPTQBit=2, Model Architecture=Qwen2-VL-72B2026.05 | 67.1 | |
| QuIP#Bit=2, Model Architecture=InternVL2-26B2026.05 | 67 | |
| FP16Bit=FP16, Model Architecture=LLaVA-onevision-7B2026.05 | 66.9 | |
| MGVQBit=2, Model Architecture=InternVL2-8B2026.05 | 66.7 | |
| MBQBit=2, Model Architecture=LLaVA-onevision-72B2026.05 | 66.5 | |
| MBQBit=2, Model Architecture=InternVL2-26B2026.05 | 66.5 | |
| AWQBit=2, Model Architecture=Qwen2-VL-72B2026.05 | 66.3 | |
| VPTQBit=2, Model Architecture=InternVL2-26B2026.05 | 66.3 | |
| MGVQBit=3, Model Architecture=LLaVA-onevision-7B2026.05 | 65.8 | |
| RTNBit=3, Model Architecture=Qwen2-VL-7B2026.05 | 65.4 | |
| VPTQBit=3, Model Architecture=LLaVA-onevision-7B2026.05 | 65.3 | |
| MBQBit=2, Model Architecture=Qwen2-VL-7B2026.05 | 65.1 | |
| MBQBit=3, Model Architecture=LLaVA-onevision-7B2026.05 | 64.8 | |
| RTNBit=2, Model Architecture=Qwen2-VL-72B2026.05 | 64 | |
| GPTQBit=3, Model Architecture=LLaVA-onevision-7B2026.05 | 63.4 | |
| QuIP#Bit=2, Model Architecture=InternVL2-8B2026.05 | 62.4 | |
| MGVQBit=2, Model Architecture=LLaVA-onevision-7B2026.05 | 61.5 | |
| AWQBit=2, Model Architecture=InternVL2-26B2026.05 | 61.4 | |
| VPTQBit=2, Model Architecture=InternVL2-8B2026.05 | 61.3 | |
| AWQBit=3, Model Architecture=LLaVA-onevision-7B2026.05 | 60.4 | |
| GPTQBit=2, Model Architecture=InternVL2-26B2026.05 | 60.2 | |
| MBQBit=2, Model Architecture=LLaVA-onevision-7B2026.05 | 59.6 | |
| VPTQBit=2, Model Architecture=LLaVA-onevision-7B2026.05 | 59.6 | |
| RTNBit=2, Model Architecture=InternVL2-26B2026.05 | 59.4 | |
| MBQBit=2, Model Architecture=InternVL2-8B2026.05 | 59.1 | |
| QuIP#Bit=2, Model Architecture=LLaVA-onevision-7B2026.05 | 58.9 | |
| AWQBit=3, Model Architecture=LLaVA-onevision-72B2026.05 | 55.1 | |
| AWQBit=2, Model Architecture=Qwen2-VL-7B2026.05 | 53 | |
| GPTQBit=2, Model Architecture=Qwen2-VL-7B2026.05 | 52.5 | |
| RTNBit=2, Model Architecture=Qwen2-VL-7B2026.05 | 51.8 | |
| GPTQBit=2, Model Architecture=LLaVA-onevision-72B2026.05 | 50.5 | |
| AWQBit=2, Model Architecture=InternVL2-8B2026.05 | 48.5 | |
| RTNBit=3, Model Architecture=LLaVA-onevision-7B2026.05 | 47.9 | |
| GPTQBit=2, Model Architecture=InternVL2-8B2026.05 | 46.6 | |
| RTNBit=2, Model Architecture=InternVL2-8B2026.05 | 44.3 | |
| RTNBit=2, Model Architecture=LLaVA-onevision-72B2026.05 | 42.2 | |
| AWQBit=2, Model Architecture=LLaVA-onevision-72B2026.05 | 42.2 | |
| AWQBit=2, Model Architecture=LLaVA-onevision-7B2026.05 | 40.4 | |
| GPTQBit=2, Model Architecture=LLaVA-onevision-7B2026.05 | 38 | |
| RTNBit=2, Model Architecture=LLaVA-onevision-7B2026.05 | 25.5 |