Bengaluru‑based AI startup Sarvam AI announced on Wednesday that it will open‑source the foundational models it is developing under India’s $1.2 billion IndiaAI Mission, co‑founder Vivek Raghavan told The Economic Times.
Sarvam AI was selected in the mission’s first phase to build an indigenous large‑language model, joining a cohort of 43 proposals dedicated to LLM development out of more than 500 submissions, IndiaAI Mission CEO Abhishek Singh said in a virtual address at an IIIT‑Delhi open‑source software meet.
Singh also confirmed that the government‑sponsored LLMs, including Sarvam’s, will be made publicly available, underscoring New Delhi’s commitment to sovereign, culturally relevant AI.
The company, which secured the mission’s largest single subsidy of ₹98.68 crore to access 4,096 Nvidia H100 GPUs over six months, will release its models under permissive licences to foster broad adoption and innovation.
The decision follows public pressure by industry figures such as Paras Chopra, founder of Wingify, who questioned why startups backed by private funds were free to open‑source their work while Sarvam’s taxpayers‑subsidised effort remained closed.
In response, co‑founder Pratyush Kumar clarified that the IndiaAI Mission’s funding is structured as a compute‑for‑equity arrangement and highlighted Sarvam’s pledge to build public‑interest use cases and drive down inference costs within India.
Launched with cabinet approval in March 2024 and targeting the procurement of over 10,000 GPUs, the IndiaAI Mission incentivises startups like Sarvam AI, Gnani, GAN and SoKet AI Labs with capital and infrastructure support. The programme aims to position India as a global AI leader by fostering homegrown foundational models that reflect the country’s linguistic and cultural diversity.
Sunil Gupta, CEO of Yotta Data Services, noted at the AI4India Summit in Bengaluru on June 28, 2025, that the mission’s emphasis on LLMs signals a strategic pivot towards building an AI ecosystem that is both cutting‑edge and rooted in local needs. By open‑sourcing these government‑backed models, New Delhi seeks to catalyse research, reduce dependencies on foreign technologies and accelerate AI adoption across industry and academia.
