Xiaomi has unveiled its new MiMo-V2.6 series of multimodal AI models, consisting of the flagship MiMo-V2.6 Pro and the optimized MiMo-V2.6-Flash. Both models are capable of processing text, code, images, and video content.
The MiMo-V2.6 Pro model has secured the top position in the Artificial Analysis Intelligence Index among open-source models. It demonstrates competitive performance against industry leaders such as OpenAI, DeepSeek, Anthropic, and Google.
Xiaomi has released the model weights on HuggingFace and OpenRouter, alongside over 7,000 reinforcement learning environments. The company also provided API access at the same pricing tier as the previous V2.5 Pro version.
The models feature spatial reasoning capabilities, demonstrated through the Vibe World project, which includes generating Blender models and controlling a Franka Panda robotic arm. Additionally, the AI can generate music based on user prompts.
