IFM released six open K2 Horizon AI models, a user reports

A user reports that the Institute of Foundation Models (IFM) released K2 Horizon, six open AI models ranging from 0.9 billion to 375 billion parameters, spanning on-device models through larger models for reasoning, coding and AI agents. The post says each model was trained on roughly 20 trillion tokens and describes releases of weights, training code, configurations, data or detailed data recipes, checkpoints, logs and evaluations. It also says IFM is releasing xLLM, the training infrastructure used to build the models, and its full agent post-training codebase, including the reinforcement-learning pipeline.