Multimodal Understanding on RealWorldQA
78RWQA ScoreInternVL3-78B
Evaluation Results
| Method | Links | |
|---|---|---|
| InternVL3-78Bnumber of parameters=78B2026.02 | 78 | |
| OmniInference scheme=no-thinking2026.04 | 76 | |
| InternVL3-38Bnumber of parameters=38B2026.02 | 75.6 | |
| Qwen3-VL-30B-A3B-InstructInference scheme=no-thinking2026.04 | 73.7 | |
| InternVL3.5-30B-A3BInference scheme=no-thinking2026.04 | 72.3 | |
| Innovator-VLvariant=8B-Instruct, parameters=8B2026.01 | 71.5 | |
| Innovator-VLvariant=8B-Thinking, parameters=8B2026.01 | 70.98 | |
| TPRU-32Bnumber of parameters=32B2026.02 | 69.8 | |
| TPRU-7Bnumber of parameters=7B2026.02 | 69.54 | |
| MiniCPM-Vversion=4.5, parameters=8B2026.01 | 69.28 | |
| Qwen3-VLparameters=8B2026.01 | 69.02 | |
| Qwen2.5-VLParams=7B2026.04 | 69 | |
| Qwen3-VLParams=8B2026.04 | 69 | |
| Qwen2.5-VL-32Bnumber of parameters=32B2026.02 | 68.89 | |
| Qwen2.5-VL-7Bnumber of parameters=7B2026.02 | 68.1 | |
| InternVL3Params=8B2026.04 | 67.8 | |
| LLaVA-OVversion=1.5, parameters=8B2026.01 | 67.71 | |
| MiMo-VLtraining=7B-SFT, parameters=7B2026.01 | 67.19 | |
| TPRU-3Bnumber of parameters=3B2026.02 | 66.54 | |
| MiMo-VLtraining=7B-RL, parameters=7B2026.01 | 66.01 | |
| Qwen2.5-VL-3Bnumber of parameters=3B2026.02 | 65.36 | |
| InternVL3.5parameters=8B2026.01 | 64.71 | |
| Intern-S1variant=mini, parameters=9B2026.01 | 63.66 | |
| InternVL3.5Params=8B2026.04 | 63 | |
| GPT-5 Nano2026.04 | 60.7 | |
| EMU3Backbone=8B from scratch, Resolution=512, Visual Input Type=Discrete2024.12 | 57.4 | |
| TokenFlow-XLBackbone=Qwen-2.5-14B, Resolution=384, Visual Input Type=Discrete2024.12 | 56.6 | |
| LLaVA-1.5Backbone=Vicuna-13B, Resolution=336, Visual Input Type=Continuous2024.12 | 55.3 | |
| ShareGPT4VBackbone=Vicuna-7B, Resolution=336, Visual Input Type=Continuous2024.12 | 54.9 | |
| TokenFlow-XLBackbone=Vicuna-13B, Resolution=384, Visual Input Type=Discrete2024.12 | 53.7 | |
| UniDriveVLAParams=8B2026.04 | 49.9 | |
| TokenFlow-BBackbone=Vicuna-13B, Resolution=224, Visual Input Type=Discrete2024.12 | 49.4 | |
| TokenFlow-LBackbone=Vicuna-13B, Resolution=256, Visual Input Type=Discrete2024.12 | 49.2 |