OPEN-1B is out — a fully auditable LLM whose training steps can be replayed and verified bit for bit. Preprint · announcement.