The Indian government is making significant strides toward developing its own indigenous Large Language Model (LLM), with a vision to compete with global AI giants like ChatGPT and DeepSeek. This initiative, part of the IndiaAI Mission, marks a major step toward ensuring India's sovereignty in the rapidly evolving AI sector. The Government of India has officially selected Sarvam, an Indian AI company, to build this sovereign LLM.
Sarvam, renowned for its capabilities in developing AI models tailored for Indian languages, will receive dedicated computing resources to create a homegrown foundational model. This LLM will be designed to have advanced reasoning capabilities, be voice-centric, and support multiple Indian languages. Additionally, it will be engineered for secure, large-scale deployment, emphasizing the importance of building AI infrastructure within India. As mentioned in the official statement, the model will be “built, deployed, and optimized in India,” relying on domestic infrastructure and the talent pool within the country.
The move to develop a sovereign LLM is rooted in India's desire for strategic independence in AI, fostering innovation, and securing leadership in the global AI landscape. The government’s commitment to the project was underscored by Union Minister of Electronics and IT Ashwini Vaishnaw’s earlier announcement, which highlighted a 10-month timeline for a competitive AI model.
Sarvam, which has already demonstrated success in developing models that cater to Indian languages, aims to offer solutions that are deeply integrated into India’s cultural and linguistic landscape. Dr. Pratyush Kumar, Co-founder of Sarvam, emphasized the company's mission to build an AI ecosystem within India, enabling local developers to create impactful solutions. The company’s models are expected to rival global counterparts, with the government expressing confidence in their competitiveness.
Sarvam's AI initiative includes three distinct models to cater to varying needs:
-
Sarvam-Large: Designed for advanced reasoning and generation tasks.
-
Sarvam-Small: Geared for real-time interactive applications.
-
Sarvam-Edge: Focused on compact, on-device tasks, facilitating local processing without relying on cloud infrastructure.
The company is also collaborating with AI4Bharat at IIT Madras, a leader in Indian language AI research, to further enhance the capabilities of these models. The goal is to create multi-modal, multi-scale foundation models that can serve a wide range of applications across India. This project not only promises to make AI more accessible to Indian citizens but also ensures that sensitive data remains within the country’s borders, avoiding the risks associated with sharing data with foreign entities.
In the words of Dr. Vivek Raghavan, Co-founder of Sarvam, the company views this responsibility as a significant step in building critical national AI infrastructure. The outcome of this initiative is expected to unlock a range of applications for both citizens and businesses, fostering an AI-driven future for India. The model will be optimized for real-world use, with a focus on making AI interactions more relatable and familiar to Indian users, unlike the typically foreign-feeling AI systems in use today.
This bold step aligns with the Indian government’s broader push to develop cutting-edge technology within its own borders and to provide a sovereign alternative in a world where AI is becoming increasingly integral to both daily life and business operations.
