Indian AI Startup Sarvam AI to Develop Trillion-Parameter Model
Indian AI startup Sarvam AI has announced plans to develop a trillion-parameter AI model within the next six months to foster domestic AI sovereignty. Alongside this, the company launched the Bulbul V4 text-to-speech model and the Saaras V4 speech recognition model, while announcing a new US office to attract Indian talent back home. This initiative represents a significant step toward India's "sovereign AI" goals, aiming to keep data and token consumption local while reducing financial dependency on foreign AI giants. If successful, the trillion-parameter model could compete globally in specialized fields like coding, cybersecurity, and scientific research. While the trillion-parameter model is currently an announcement of intent rather than a finished product, Sarvam AI claims its existing 105-billion parameter model and voice models already deliver superior cost-performance compared to global competitors.
## BACKGROUND
Sovereign AI refers to national or regional efforts to build independent AI infrastructure, models, and data controls to reduce critical reliance on foreign technology providers. Sarvam AI is a prominent Indian startup focused on building full-stack sovereign AI systems tailored to India's diverse languages and local needs.