The Indian government has chosen Bengaluru-based startup Sarvam to lead the development of the nation’s first indigenous Large Language Model (LLM) as part of the IndiaAI Mission.
Sarvam is working on three model variants-
Sarvam-Large: designed for advanced reasoning
Sarvam-Small: tailored for real-time use cases
Sarvam-Edge: optimized for compact, on-device operations
As part of the initiative, the company will be granted access to 4,000 Graphics Processing Units (GPUs) over a six-month period to build a 70-billion-parameter AI model.
A senior official said that the model is not expected to be open-sourced, but will be fine-tuned particularly for Indian languages.
IT Minister Ashwini Vaishnaw said: “This (Sarvam’s) model will have 70 billion parameters and many innovations in programming as well as engineering. With these innovations, a 70 billion parameter (model) can compete with some of the best in the world.”
Dr. Vivek Raghavan, Co-founder of Sarvam, said, “We are humbled by the responsibility bestowed upon us to build India’s sovereign model, and we are ready to build AI that reaches every corner of the country. This is a crucial step toward building critical national AI infrastructure. Our goal is to build multi-modal, multi-scale foundation models from scratch. When we do, a universe of applications unfolds. For citizens, this means interacting with AI that feels familiar, not foreign. For enterprises, this means unlocking intelligence without sending their data beyond borders.”
“We are deeply grateful to the Government of India for its vision and support in advancing AI,” said Dr. Pratyush Kumar, Co-founder of Sarvam. “Building an AI ecosystem for India has always been core to Sarvam’s mission, where our research, technology, and models empower builders to create solutions for the country. As part of the Sovereign LLM proposal, we are developing three model variants: Sarvam-Large for advanced reasoning and generation, Sarvam-Small for real-time interactive applications, and Sarvam-Edge for compact on-device tasks. We are collaborating with AI4Bharat at IIT Madras, a leader in Indian language AI research, to build these models. Driving this effort is a best-in-class team at Sarvam that understands the depth and complexity of AI development like few others.”
This development unfolds against the backdrop of DeepSeek’s rapid ascent, a low-cost foundational model from China.
Soumyarendra Barik from Indian Express notes that- being open source, and reportedly developed at a fraction of the cost of its U.S. counterparts, DeepSeek’s emergence has disrupted the AI space. Its R1 model, trained on less powerful GPUs than those used by firms like OpenAI, contributed to a dip in Nvidia’s stock value, highlighting shifting dynamics in the AI hardware and model ecosystem.
According to a press statement from the company, the sovereign model will be built, deployed, and optimized in India, using local infrastructure and developed by a new generation of Indian talent. This initiative aims to promote strategic autonomy, accelerate domestic innovation, and secure India’s leadership in AI for the long term.
