MiMo-VL

model paper
Xiaomi
Open-source vision-language model that outperformed models ten times its size (like Qwen-72B) in multimodal reasoning. Includes SFT and RL variants.
HuggingFacePaper (arXiv)GitHub

Outputs 2

MiMo-VL-7B

model
HuggingFaceGitHub
Architecture DENSE
Parameters 7B

MiMo-VL: From Pre-training to Post-training

paper

Technical report on achieving SOTA multimodal reasoning at the 7B scale.

Paper (arXiv)

arXiv: 2506.03569

multimodalreasoningopen-weight