Multimodal Understanding (Chinese) on MMBench Chinese
91AccuracyInternVL3-8B-Masters
Evaluation Results
| Method | Links | |
|---|---|---|
| InternVL3-8B-MastersSize=8B2025.12 | 91 | |
| Qwen3-VL-8B-MastersSize=8B2025.12 | 89.9 | |
| Qwen2.5-VL-7B-MastersSize=7B2025.12 | 89.5 | |
| Qwen3-VL-4B-MastersSize=4B2025.12 | 89.3 | |
| InternVL3.5-8B-MastersSize=8B2025.12 | 88.3 | |
| Qwen3-VL-235BToken Retention Ratio=100%2026.05 | 88.1 | |
| F3AToken Retention Ratio=60%2026.05 | 86.76 | |
| VisionZipToken Retention Ratio=60%2026.05 | 86.71 | |
| Qwen3-VL-8BSize=8B2025.12 | 86.5 | |
| DivPruneToken Retention Ratio=60%2026.05 | 86.38 | |
| Qwen2.5-VL-3B-MastersSize=3B2025.12 | 86.3 | |
| InternVL3.5-2B-MastersSize=2B2025.12 | 86.3 | |
| F3AToken Retention Ratio=40%2026.05 | 86.14 | |
| InternVL3.5-4BSize=4B2025.12 | 86.1 | |
| CDPrunerToken Retention Ratio=60%2026.05 | 86.01 | |
| InternVL3.5-4B-MastersSize=4B2025.12 | 86 | |
| InternVL3.5-8BSize=8B2025.12 | 85.9 | |
| VisionZipToken Retention Ratio=40%2026.05 | 85.7 | |
| FastVToken Retention Ratio=60%2026.05 | 85.42 | |
| DivPruneToken Retention Ratio=40%2026.05 | 85.36 | |
| InternVL3-2B-MastersSize=2B2025.12 | 85.2 | |
| CDPrunerToken Retention Ratio=40%2026.05 | 85.12 | |
| FastVToken Retention Ratio=40%2026.05 | 84.66 | |
| Qwen3-VL-4BSize=4B2025.12 | 84.3 | |
| Qwen3-VL-2B-MastersSize=2B2025.12 | 84.1 | |
| F3AToken Retention Ratio=20%2026.05 | 84.06 | |
| Qwen2.5-VL-7BSize=7B2025.12 | 83.4 | |
| CDPrunerToken Retention Ratio=20%2026.05 | 82.96 | |
| VisionZipToken Retention Ratio=20%2026.05 | 82.73 | |
| DivPruneToken Retention Ratio=20%2026.05 | 82.41 | |
| InternVL3-8BSize=8B2025.12 | 82.2 | |
| MiMo-VL-8BSize=8B2025.12 | 82 | |
| InternVL3.5-2BSize=2B2025.12 | 81.8 | |
| Qwen3-VL-2BSize=2B2025.12 | 81.4 | |
| LLaVA-OneVision-1.5-8BSize=8B2025.12 | 81 | |
| FastVToken Retention Ratio=20%2026.05 | 80.74 | |
| InternVL3-2BSize=2B2025.12 | 78.4 | |
| Qwen2.5-VL-3BSize=3B2025.12 | 78.1 | |
| InternLM-XComposer2Number of Parameters=7B2024.01 | 77.6 | |
| LLaVA-OneVision-1.5-4BSize=4B2025.12 | 76.9 | |
| GPT-4V2024.01 | 74.4 | |
| Gemini-Pro2024.01 | 74.3 | |
| L-Int2Number of Parameters=20B2024.01 | 73.7 | |
| Qwen-VL-Plus2024.01 | 70.7 | |
| LLaVA-1.5 + MMInstructLLM=Vicuna-13B2024.07 | 68 | |
| Phi-3.5-Vision-4BSize=4B2025.12 | 66.1 | |
| LLaVA-1.5 + MMInstructLLM=Vicuna-7B2024.07 | 65.2 | |
| LLaVA-1.5-13BLLM=Vicuna-1.5-13B, Resolution=3362024.06 | 63.6 | |
| LLaVA-1.5LLM=Vicuna-13B2024.07 | 63.6 | |
| LLaVA-Instruct-13BLLM=Vicuna-1.5-13B, Resolution=3362024.06 | 62.5 | |
| LLaVA-Instruct-7BLLM=Vicuna-1.5-7B, Resolution=3362024.06 | 60 | |
| LLaVA-1.5-7BLLM=Vicuna-1.5-7B, Resolution=3362024.06 | 58.3 | |
| LLaVA-1.5LLM=Vicuna-7B2024.07 | 58.3 | |
| Qwen-VL-ChatLLM=Qwen-7B, Resolution=4482024.06 | 56.7 | |
| Qwen-VL-chatLLM=Qwen-7B2024.07 | 56.7 | |
| IDEFICS-80BLLM=LLaMA-65B, Resolution=2242024.06 | 38.1 | |
| IDEFICS-80BLLM=LLaMA-65B2024.07 | 38.1 | |
| IDEFICS-9BLLM=LLaMA-7B, Resolution=2242024.06 | 25.2 | |
| IDEFICS-9BLLM=LLaMA-7B2024.07 | 25.2 | |
| InstructBLIP-7BLLM=Vicuna-7B, Resolution=2242024.06 | 23.7 | |
| InstructBLIPLLM=Vicuna-7B2024.07 | 23.7 | |
| Qwen-VLLLM=Qwen-7B, Resolution=4482024.06 | 7.4 | |
| Qwen-VLLLM=Qwen-7B2024.07 | 7.4 |