Particle.news

Download on the App Store

IndiaAI Mission Taps Sarvam AI to Build Nation’s First Sovereign Large Language Model

The Bengaluru-based startup begins development of a 70-billion-parameter, multilingual AI model using 4,000 GPUs, focusing on strategic autonomy and inclusive access.

Image

Overview

  • The Government of India has selected Sarvam AI to develop its first indigenous large language model under the ₹10,370-crore IndiaAI Mission.
  • Sarvam AI is building a 70-billion-parameter model optimized for voice interaction and fluency in Indian languages, with three variants: Sarvam-Large, Sarvam-Small, and Sarvam-Edge.
  • The project leverages 4,000 GPUs for six months and domestic infrastructure to ensure the model is developed entirely within India.
  • Key partnerships include AI4Bharat at IIT Madras and GPU-as-a-service providers like Jio, Yotta, and Tata Communications to establish central compute infrastructure.
  • The initiative aims to foster technological sovereignty, democratize AI access for rural and non-tech users, and position India as a global AI innovator.