Sarvam AI, a Bengaluru-based AI startup, has been selected by the Indian government to spearhead the development of India's first independent large language model (LLM). This initiative falls under the IndiaAI Mission, a program with an outlay of ₹10,371 crore, aimed at fostering AI innovation and building a robust AI ecosystem in the country. This selection marks a significant step towards India's self-reliance in AI and its ambition to compete with global AI leaders.
The IndiaAI Mission and Sarvam's Role
The IndiaAI Mission underscores the government's commitment to boosting India's AI capabilities by investing in compute capacity, research, talent development, datasets, and startups. A key component of this mission is the IndiaAI Compute Capacity initiative, which aims to make AI infrastructure accessible and affordable to startups, researchers, and institutions. As part of this initiative, Sarvam AI will receive dedicated compute resources to build its LLM from the ground up. This includes access to 400 high-end GPUs for six months, which are essential for training large AI models. The commercial value of this access is estimated to be approximately INR 220 Cr.
Building a Sovereign AI Model
Sarvam AI's mandate is to create a sovereign LLM that is fluent in Indian languages, capable of reasoning, and designed for voice-based interactions. The model will be fully developed, deployed, and optimized within India, utilizing local infrastructure and talent. This approach ensures data security, strategic autonomy, and promotes domestic innovation. The company aims to build multi-modal, multi-scale foundation models from scratch.
Dr. Vivek Raghavan, Co-founder of Sarvam, emphasized the importance of building AI that feels familiar to Indian citizens and allows enterprises to leverage AI without sending data abroad. Union Minister for Electronics & Information Technology, Railways, and Information & Broadcasting, Shri Ashwini Vaishnaw, expressed confidence that Sarvam's models will be competitive with global models.
Model Variants and Capabilities
Sarvam AI is developing three variants of the LLM to cater to different needs:
The model is expected to have 70 billion parameters, putting it on par with some of the best models globally. It will support both Indian languages and English. The company is collaborating with AI4Bharat at IIT Madras to develop these models, leveraging their expertise in Indian language AI research.
Sarvam AI: A Rising Force
Founded in 2023 by Vivek Raghavan and Pratyush Kumar, Sarvam AI has quickly established itself as a key player in India's AI landscape. The company secured $41 million in Series A funding in December 2023, led by Lightspeed Venture Partners, with participation from Peak XV Partners and Khosla Ventures. This funding has enabled Sarvam to develop a full-stack GenAI platform that includes various products such as Sarvam Agents, Sarvam 2B, Shuka 1.0, Sarvam Models, and A1. Sarvam's platform empowers businesses and developers to build and test generative AI applications, particularly voice-first interfaces, which are crucial for bridging digital access gaps in India.
Sarvam AI's existing models have demonstrated strong performance in Indian languages, outperforming leading frontier models on various benchmarks. The company's clients include UIDAI, Urban Company, NITI Aayog, and the Ministry of Skill Development.
Towards an AI-Driven Future for India
Sarvam AI's selection to build India's first sovereign LLM is a pivotal moment for the nation's AI ecosystem. This initiative will enable India to reduce its dependence on foreign AI models and create AI solutions that are tailored to its unique linguistic and cultural needs. By fostering domestic innovation and building a strong AI infrastructure, India aims to secure its position as a leader in the global AI landscape.
