Multimodal Understanding on MMBench CN
88.5AccuracyInternVL2.5-78B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| InternVL2.5-78BModel Name=InternVL2.5-78B2024.12 | 88.5 | — | — | — | |
| Qwen2-VL-72BModel Name=Qwen2-VL-72B2024.12 | 86.6 | — | — | — | |
| InternVL2-40BModel Name=InternVL2-40B2024.12 | 86.5 | — | — | — | |
| InternVL2.5-38BModel Name=InternVL2.5-38B2024.12 | 86.3 | — | — | — | |
| InternVL2-Llama3-76BModel Name=InternVL2-Llama3-76B2024.12 | 86.3 | — | — | — | |
| VLsI-7BParameters=7B2024.12 | 85.5 | — | — | — | |
| InternVL2.5-26BModel Name=InternVL2.5-26B2024.12 | 85.5 | — | — | — | |
| LLaVA-OneVision-72BModel Name=LLaVA-OneVision-72B2024.12 | 85.3 | — | — | — | |
| InternVL3.5-38BRatio=100%, Backbone=InternVL3.5-38B2026.05 | 84.96 | — | — | — | |
| F3ARatio=60%, Backbone=InternVL3.5-38B2026.05 | 84.71 | — | — | — | |
| Phantom-7BParameters=7B2024.12 | 84.7 | — | — | — | |
| CDPrunerRatio=60%, Backbone=InternVL3.5-38B2026.05 | 84.45 | — | — | — | |
| DivPruneRatio=60%, Backbone=InternVL3.5-38B2026.05 | 84.45 | — | — | — | |
| VisionZipRatio=60%, Backbone=InternVL3.5-38B2026.05 | 84.17 | — | — | — | |
| FastVRatio=60%, Backbone=InternVL3.5-38B2026.05 | 83.54 | — | — | — | |
| DivPruneRatio=40%, Backbone=InternVL3.5-38B2026.05 | 83.52 | — | — | — | |
| Claude-3.5-SonnetModel Name=Claude-3.5-Sonnet2024.12 | 83.5 | — | — | — | |
| Qwen2.5-VL-7BToken Budget=Full2026.01 | 83.4 | — | — | — | |
| Qwen2.5-VL-7B2025.12 | 83.4 | — | — | — | |
| F3ARatio=40%, Backbone=InternVL3.5-38B2026.05 | 83.35 | — | — | — | |
| VisionZipRatio=40%, Backbone=InternVL3.5-38B2026.05 | 82.91 | — | — | — | |
| FastVRatio=40%, Backbone=InternVL3.5-38B2026.05 | 82.69 | — | — | — | |
| InternVL2.5-8B2025.12 | 82.6 | — | — | — | |
| InternVL2.5-8BModel Name=InternVL2.5-8B2024.12 | 82.6 | — | — | — | |
| CDPrunerRatio=40%, Backbone=InternVL3.5-38B2026.05 | 82.58 | — | — | — | |
| GPT-4o-20240513Model Name=GPT-4o-202405132024.12 | 82.1 | — | — | — | |
| InternVL-Chat-V1.5Model Name=InternVL-Chat-V1.52024.12 | 82 | — | — | — | |
| InternVL2-26BModel Name=InternVL2-26B2024.12 | 82 | — | — | — | |
| VisionTrimToken Budget=~1/32026.01 | 81.8 | — | — | — | |
| F3ARatio=20%, Backbone=InternVL3.5-38B2026.05 | 81.29 | — | — | — | |
| InternVL2-8BParameters=8B2024.12 | 81.2 | — | — | — | |
| TroL-7BParameters=7B2024.12 | 81.2 | — | — | — | |
| TroL-7BParameters=7B2024.06 | 81.2 | — | — | — | |
| InternVL2-8B2025.12 | 81.2 | — | — | — | |
| InternVL2-8BModel Name=InternVL2-8B2024.12 | 81.2 | — | — | — | |
| VisionZipRatio=20%, Backbone=InternVL3.5-38B2026.05 | 81.01 | — | — | — | |
| CDPrunerRatio=20%, Backbone=InternVL3.5-38B2026.05 | 80.88 | — | — | — | |
| CoT4DET-7B2025.12 | 80.7 | — | — | — | |
| Qwen2-VL-7BParameters=7B2024.12 | 80.5 | — | — | — | |
| Qwen2-VL-7B2025.12 | 80.5 | — | — | — | |
| Qwen2-VL-7BModel Name=Qwen2-VL-7B2024.12 | 80.5 | — | — | — | |
| FastVRatio=20%, Backbone=InternVL3.5-38B2026.05 | 80.39 | — | — | — | |
| SparseVLMToken Budget=~1/32026.01 | 80.3 | — | — | — | |
| GPT-4VModel Name=GPT-4V2024.12 | 80.2 | — | — | — | |
| DivPruneRatio=20%, Backbone=InternVL3.5-38B2026.05 | 80.03 | — | — | — | |
| InternVL2.5-4BModel Name=InternVL2.5-4B2024.12 | 79.3 | — | — | — | |
| MiniCPM-V2.6Model Name=MiniCPM-V2.62024.12 | 79.3 | — | — | — | |
| Cambrian-34BModel Name=Cambrian-34B2024.12 | 79.2 | — | — | — | |
| LLaVA-NeXTLLM=Hermes-Yi-34B, PT=558K, SFT=760K, #IMG=28802024.04 | 79 | — | — | — | |
| MoVALLM=Hermes-Yi-34B, PT=15M, SFT=1.6M, #IMG=5762024.04 | 79 | — | — | — | |
| VLSI-2BParameters=2B2024.12 | 78.8 | — | — | — | |
| InternLM-XComposer2LLM=InternLM2-7B2024.01 | 77.6 | — | — | — | |
| TroL-3.8BParameters=3.8B2024.12 | 77.1 | — | — | — | |
| Phantom-3.8BParameters=3.8B2024.12 | 77.1 | — | — | — | |
| TroL-3.8BModel Size=3.8B2024.09 | 77.1 | — | — | — | |
| Phantom-3.8BModel Size=3.8B2024.09 | 77.1 | — | — | — | |
| PDropToken Budget=~1/32026.01 | 77 | — | — | — | |
| Qwen-VL-MAXLLM Size=Unk2024.03 | 76.4 | — | — | — | |
| FastVToken Budget=~1/32026.01 | 76.3 | — | — | — | |
| Phantom-1.8BParameters=1.8B2024.12 | 75.1 | — | — | — | |
| Phantom-1.8BModel Size=1.8B2024.09 | 75.1 | — | — | — | |
| Qwen-VL-Max2024.04 | 75.1 | — | — | — | |
| GPT-4VLLM Size=Unk2024.03 | 74.7 | — | — | — | |
| Gemini-Pro2024.04 | 74.3 | — | — | — | |
| MiniCPM-V2.5-8BParameters=8B2024.12 | 74.2 | — | — | — | |
| TroL-1.8BParameters=1.8B2024.12 | 74.1 | — | — | — | |
| TroL-1.8BModel Size=1.8B2024.09 | 74.1 | — | — | — | |
| Gemini ProLLM Size=Unk2024.03 | 74 | — | — | — | |
| InternVL2-4BParameters=4B2024.12 | 73.9 | — | — | — | |
| GPT-4V2024.04 | 73.9 | — | — | — | |
| InternVL2-4BModel Name=InternVL2-4B2024.12 | 73.9 | — | — | — | |
| Gemini-1.5-ProModel Name=Gemini-1.5-Pro2024.12 | 73.8 | — | — | — | |
| LLaVA-XTunerLLM=InternLM2-20B2024.01 | 73.7 | — | — | — | |
| Qwen2-VL-2BParameters=2B2024.12 | 73.5 | — | — | — | |
| Qwen2-VL-2BModel Name=Qwen2-VL-2B2024.12 | 73.5 | — | — | — | |
| DeepSeek-VLLLM Size=7B2024.03 | 72.8 | — | — | — | |
| InternLM-XCLLM=InternLM-7B2024.01 | 72.4 | — | — | — | |
| InternLM-XC-7BParameters=7B2024.06 | 72.4 | — | — | — | |
| InternVL2.5-2BModel Name=InternVL2.5-2B2024.12 | 71.9 | — | — | — | |
| VILA2-8BParameters=8B2024.12 | 71.7 | — | — | — | |
| DPVR-PCTrainable=202M, evaluation=3-seed average2026.06 | 71.5 | — | — | — | |
| DPVR-KVTrainable=202M2026.06 | 71.5 | — | — | — | |
| InternVL2-2BParameters=2B2024.12 | 70.9 | — | — | — | |
| InternVL2-2BModel Name=InternVL2-2B2024.12 | 70.9 | — | — | — | |
| FastVTrainable=0, mode=inference-only2026.06 | 70.6 | — | — | — | |
| LoRATrainable=80M, r=642026.06 | 70.5 | — | — | — | |
| LoRATrainable=20M, r=162026.06 | 70.3 | — | — | — | |
| DPVR-LFTrainable=202M, evaluation=3-seed average2026.06 | 70.3 | — | — | — | |
| Phantom-0.5BModel Size=0.5B2024.09 | 70.1 | — | — | — | |
| VILA1.5-8BParameters=8B2024.12 | 69.9 | — | — | — | |
| Qwen-VL-PlusLLM Size=Unk2024.03 | 69.6 | — | — | — | |
| Vanilla LLaVA-1.5-7BTrainable=02026.06 | 69.3 | — | — | — | |
| Yi-VLLLM Size=6B2024.03 | 68.9 | — | — | — | |
| LLaVA-NeXT-13BParameters=13B2024.12 | 68.5 | — | — | — | |
| LLaVA-NeXT-13BParameters=13B2024.06 | 68.5 | — | — | — | |
| Qwen-VL-Plus2024.04 | 68 | — | — | — | |
| MoVALLM=Llama3-8B, PT=15M, SFT=1.6M, #IMG=5762024.04 | 67.7 | — | — | — | |
| Monkey-10BLLM=Qwen-7B2024.01 | 67.5 | — | — | — | |
| Monkey-10BParameters=10B2024.06 | 67.5 | — | — | — | |
| MiniCPM-V2-2.8BParameters=2.8B2024.12 | 66.5 | — | — | — |