TLDR: Indian AI company Sarvam has been selected by the Government of India to develop the nation’s first sovereign Large Language Model (LLM) as part of the ambitious IndiaAI Mission. This initiative aims to create an indigenous, multi-modal LLM fluent in Indian languages, designed for population-scale deployment, and ensuring data sovereignty. The project is backed by significant government support, including the provision of 4,000 high-end GPUs.
In a significant stride towards technological self-reliance, Indian artificial intelligence firm Sarvam has been officially selected by the Government of India to spearhead the development of the country’s inaugural sovereign Large Language Model (LLM). This landmark announcement, made in April 2025, positions Sarvam at the forefront of the IndiaAI Mission, a strategic initiative aimed at fostering indigenous AI capabilities and securing India’s leadership in the global AI landscape.
The initiative is a cornerstone of the ‘Atmanirbhar Bharat’ (Self-Reliant India) vision, emphasizing strategic autonomy and domestic innovation. The foundational AI model will be developed and deployed entirely within India, leveraging local infrastructure and engineering talent.
Union Minister for Electronics & Information Technology, Ashwini Vaishnaw, expressed strong confidence in Sarvam’s capacity to deliver models that are competitive with global benchmarks. Dr. Vivek Raghavan, Co-founder of Sarvam, underscored the critical importance of this endeavor, stating that the goal is to build multi-modal, scalable models that cater to both citizens and enterprises while rigorously maintaining data within Indian borders.
Sarvam plans to develop three distinct variants of the LLM to address diverse application needs: Sarvam-Large, designed for advanced reasoning tasks; Sarvam-Small, optimized for real-time applications; and Sarvam-Edge, tailored for compact, on-device tasks. The company is also collaborating with AI4Bharat at IIT Madras to achieve these ambitious goals.
To bolster this crucial project, the Government of India has committed substantial support, including the provision of 4,000 high-end Graphics Processing Units (GPUs) for a period of six months. These GPUs, essential for the intensive computational demands of LLM training, will be made available through partnerships with Yotta Data Services, Tata Communications, and E2E Networks.
The broader IndiaAI Mission, launched on March 7, 2024, with an allocation of Rs 10,300 crore over five years (2024-2029), outlines several key objectives. These include developing indigenous AI capabilities for Indian languages and contexts, establishing a high-end common computing facility with 18,693 GPUs, fostering the development of indigenous GPUs within three to five years, building domestic semiconductor manufacturing capacity, and providing open access to AI infrastructure for startups, researchers, and students at nominal costs.
Sarvam’s selection followed a rigorous process that saw 67 applications nationwide, with the Bengaluru-based startup being the first to be shortlisted. While the initiative has been widely celebrated, some public discussion has emerged regarding the open-source nature of the model. Sarvam AI has clarified that openness will be prioritized once critical aspects like security, scalability, and sovereignty are firmly established.
Also Read:
- IIT Madras Professor Mitesh Khapra Honored in TIME’s 2025 AI List for Pioneering Indian Language AI
- Indian Telecom Sector Embraces AI for Intelligent Networks, Enhanced Connectivity, and Social Impact
Sarvam is a well-established entity, having raised $41 million in Series A funding in December 2024, and its founders bring significant credentials and experience in public infrastructure and AI research, including their involvement in creating digital public infrastructure like Aadhaar.
