13 4 weeks ago

India-first 1.31B base model, trained from scratch (~103B tokens + WSD cooldown on 8×H100). The completion/foundation model behind the Navya-1c chat models — not instruction-tuned. Research model - not investment advice.