Xiaomi Launches MiMo-7B AI Model, Excels in Reasoning Tasks

Xiaomi has released MiMo-7B, its first open-source AI model focused on reasoning. It outperforms major competitors like OpenAI’s o1-mini and Alibaba’s Qwen-32B in mathematical (AIME) and programming (LiveCodeBench) tests. The model achieves this by using large-scale reasoning data and reinforcement learning with a unique Test Difficulty Driven Reward system. MiMo-7B currently leads other 7B models in reasoning and learning potential. The full model and technical report are now available on Hugging Face.