IFM open-sources K2 Horizon: six models from 0.9B to 375B with full training process
rohanpaul_ai · x · 2026-09-11
Institute of Foundation Models (IFM) released K2 Horizon, six open models from 0.9B to 375B spanning on-device models to reasoning, coding and agent models.
- Fully open: weights, training code, configs, data recipes, intermediate checkpoints, training logs and evals for every model, covering pre-training through agentic post-training
- Results: the 0.9B/3.7B/7B set new SOTA in their size classes; the sparse 36B-A4B activates only 4B params/token yet approaches the dense 32B
- Uses the new Mixture-of-Value Attention (MoVA), extending MoE routing into attention layers
- Each model trained on 20T tokens; checkpoints and logs give researchers an unusually detailed development record
- Also releasing xLLM, the production training infra, plus the full agentic post-training codebase including the RL pipeline
Related event: IFM open-sources K2 Horizon models and discloses 103 benchmark cheats(7 posts)→
More from coding & agent
- Ollama lets you customize ChatGPT's model selector with local and cloud models — ollama · 2026-09-11
- ChatGPT Desktop Codex app can now use local models via Ollama 0.34 — ollama · 2026-09-11
- Inspect evals: default continuation prompt can make agents misread bad actions as approved — AdtRaghunathan · 2026-09-11
- Claude Code creator: AI-written production code should be held to a higher bar — bcherny · 2026-09-11
- Dev culture gap: some burn a week of Codex quota daily, others never tried it — yihui_indie · 2026-09-11
- Modern Context Stack: a satire site skewering the data industry's tool-buying habit — juansequeda · 2026-09-11