India's government selects local startup Sarvam to develop the country's first indigenously-built large language model.
The government has selected Bengaluru-based start-up Sarvam to build India's first indigenous artificial intelligence large language model, chosen from among 67 applicants competing for support under the government's Rs 10,370 crore IndiaAI Mission. The selection arrives amid industry disruption caused by China's low-cost foundational model DeepSeek, which has gained prominence for being open source, its accuracy, and claims of development at a fraction of the cost of US competitors. Sarvam is the first start-up approved for incentives under the IndiaAI Mission for model building, while the government continues assessing hundreds of other proposals.
In terms of government support, Sarvam will receive access to 4,000 graphics processing units for six months to build and train its model. The GPUs will be provided by companies separately selected by the government to establish AI data centres in India. The resulting model will not be open-sourced but will be fine-tuned particularly for Indian languages. According to IT Minister Ashwini Vaishnaw, "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."
As part of its LLM proposal, Sarvam is 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, according to co-founder Pratyush Kumar. The company has stated that the model will be capable of reasoning, designed for voice, and fluent in Indian languages, ready for population-scale deployment. Sarvam's model will be built, deployed, and optimized in India using local infrastructure and developed by Indian talent, with the initiative aimed at promoting strategic autonomy, accelerating domestic innovation, and securing India's long-term AI leadership.
Earlier this year, the government selected 10 companies to supply 18,693 GPUs—exceeding the IndiaAI Mission's initial aim of procuring 10,000 GPUs. The companies empaneled to provide GPU services include Jio Platforms, the Hiranandani Group-backed Yotta, Tata Communications, E2E Networks, NxtGen Datacenter, CMS Computers, Ctrls Datacenters, Locuz Enterprise Solutions, Orient Technologies, and Vensysco Technologies.
Co-founder Vivek Raghavan stated, "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."