Multimodal Understanding on MMStar (Average Score)
68.01Average ScoreLLaVA-OV
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| LLaVA-OVversion=1.5, parameters=8B2026.01 | 68.01 | — | — | — | — | — | — | |
| InternVL3.5parameters=8B2026.01 | 66.16 | — | — | — | — | — | — | |
| Innovator-VLvariant=8B-Thinking, parameters=8B2026.01 | 66.15 | — | — | — | — | — | — | |
| Innovator-VLvariant=8B-Instruct, parameters=8B2026.01 | 64.96 | — | — | — | — | — | — | |
| Intern-S1variant=mini, parameters=9B2026.01 | 63.5 | — | — | — | — | — | — | |
| LLaVA-OVType=AR, LLM Tower=Qwen2-7B2026.01 | 61.7 | — | — | — | — | — | — | |
| Qwen2-VLType=AR, LLM Tower=Qwen2-7B2026.01 | 60.7 | — | — | — | — | — | — | |
| Qwen3-VLparameters=8B2026.01 | 60.45 | — | — | — | — | — | — | |
| LLaDA-VType=Diff., LLM Tower=LLaDA-8B2026.01 | 60.1 | — | — | — | — | — | — | |
| OursType=Diff., LLM Tower=LLaDA-8B2026.01 | 59.3 | — | — | — | — | — | — | |
| MiMo-VLtraining=7B-RL, parameters=7B2026.01 | 59.25 | — | — | — | — | — | — | |
| LLaDA-V w/dllm-cacheType=Diff., LLM Tower=LLaDA-8B2026.01 | 58.3 | — | — | — | — | — | — | |
| MiMo-VLtraining=7B-SFT, parameters=7B2026.01 | 58.24 | — | — | — | — | — | — | |
| LLaMA3-VType=Diff., LLM Tower=LLaMA3-8B2026.01 | 56.5 | — | — | — | — | — | — | |
| Meteor2024.05 | 52.8 | — | — | — | — | — | — | |
| Meteor2024.05 | 52.8 | — | — | — | — | — | — | |
| MiniCPM-Vversion=4.5, parameters=8B2026.01 | 50.73 | — | — | — | — | — | — | |
| MoAI2024.05 | 48.7 | — | — | — | — | — | — | |
| MoAI2024.05 | 48.7 | — | — | — | — | — | — | |
| CoLLaVO2024.05 | 42.1 | — | — | — | — | — | — | |
| CoLLaVO2024.05 | 42.1 | — | — | — | — | — | — | |
| iGVLMBackbone=Vicuna-13B, Venue=Ours, Thpt=8.6, Δ=+3.62026.03 | 36.4 | 57.6 | 37.6 | 43.2 | 27.2 | 22.4 | 30.8 | |
| QA-ViTBackbone=Vicuna-13B, Venue=CVPR’24, Thpt=10.2, Δ=+3.12026.03 | 35.9 | 57.6 | 32.8 | 43.2 | 28.8 | 21.2 | 31.6 | |
| iGVLMBackbone=Vicuna-7B, Venue=Ours, Thpt=11.1, Δ=+4.42026.03 | 34.7 | 59.6 | 28 | 41.2 | 32.8 | 19.6 | 26.8 | |
| DyFoBackbone=Vicuna-13B, Venue=CVPR’25, Thpt=0.47, Δ=+1.82026.03 | 34.6 | 55.6 | 32 | 44.4 | 27.2 | 22 | 26.4 | |
| DyFoBackbone=Vicuna-7B, Venue=CVPR’25, Thpt=0.49, Δ=+2.72026.03 | 33 | 54.8 | 28.4 | 38 | 25.6 | 24 | 27.2 | |
| LLaVA-1.5Backbone=Vicuna-13B, Venue=CVPR’24, Thpt=9.8, Δ=–2026.03 | 32.8 | 58.8 | 28 | 41.6 | 24.4 | 18.4 | 25.6 | |
| QA-ViTBackbone=Vicuna-7B, Venue=CVPR’24, Thpt=12.9, Δ=+1.22026.03 | 31.5 | 60 | 28 | 40.4 | 25.6 | 14 | 21.2 | |
| LLaVA-1.5Backbone=Vicuna-7B, Venue=CVPR’24, Thpt=13.5, Δ=–2026.03 | 30.3 | 58.8 | 24 | 38.8 | 24 | 13.6 | 22.8 | |
| iGVLMBackbone=Qwen2.5-3B, Venue=Ours, Thpt=10.9, Δ=+4.52026.03 | 21.3 | 20 | 13.2 | 19.2 | 28.4 | 21.6 | 25.6 | |
| LLaVA-1.5Backbone=Qwen2.5-3B, Venue=CVPR’24, Thpt=12.6, Δ=–2026.03 | 16.8 | 8.4 | 8.4 | 16.4 | 24.4 | 16.8 | 26.4 |