Xiaomi's MiMo-V2.6-Pro-RL is a downloadable model checkpoint with text, image, video and audio support and a stated one-million-token context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition →. Its model card lists an MIT license and deployment configurations for running the model.
The checkpoint contains 1.02 trillion parameters but activates about 42 billion per token through a mixture-of-expertsA model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.Full definition → architecture, which selects portions of the model for each computation. Those figures describe different things: the amount of learned model data and the portion used for an individual token. The release gives researchers access to the checkpoint for their own deployment and evaluation.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
mixture-of-experts — A model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.
multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
