BharatGen Param2 17B MoE

BharatGen Param2 17B MoE, unveiled February 2026 at the India AI Impact Summit in New Delhi, is India's flagship sovereign foundation model: a 17-billion-parameter mixture-of-experts model built natively for 22 Scheduled Indian languages. Developed under the IndiaAI Mission at IIT Bombay's Technology Innovation Hub with NVIDIA partnership (NeMo, NeMo-RL, Base Command Manager), it is trained on India-centric data under the Bharat Data Sagar program and released openly on Hugging Face. It anchors a broader family including Shrutam2 (multilingual speech-to-text), Sooktam2 (text-to-speech with zero-shot voice cloning), and Patram (document processing for Indian banking, insurance, and government forms).

The honest picture: sovereign-AI announcements carry national-prestige incentives, and independent benchmark comparisons against mainstream multilingual models are still sparse. The model's real differentiator is not raw capability but first-class Indic-language coverage and public-sector alignment — governance, education, healthcare, and agriculture use cases where English-centric models underperform. For developers outside those domains, mainstream open models remain more practical.