Sarvam AI Selected to Build India’s First Homegrown Sovereign AI Model

27-04-2025

05:20 AM

1 min read

What’s in Today’s Article?

The government has selected Bengaluru-based start-up Sarvam to build the country’s first indigenous artificial intelligence (AI) large language model (LLM) amid waves made by China’s low cost model DeepSeek.

In a landmark move to bolster India’s strategic autonomy in artificial intelligence (AI), Bengaluru-based start-up Sarvam AI has been selected to build the country’s first homegrown sovereign large language model (LLM).
The project, undertaken under the government’s ambitious ₹10,370 crore IndiaAI Mission, aims to create a robust AI infrastructure fully developed, deployed, and optimized within India.
This development marks a critical step toward ensuring India’s leadership in the AI domain and promoting domestic innovation through indigenous capabilities.

The government chose Sarvam after a rigorous selection process involving 67 applicants. The start-up will receive extensive support, including access to 4,000 high-end GPUs for six months to build the model from scratch.
The GPUs will be provided through companies such as Yotta Data Services, Tata Communications, and E2E Networks, which were separately empanelled to create AI data centres in India.
The model, to be built entirely using local talent and infrastructure, will have 70 billion parameters, positioning it to compete with some of the best global AI models.
According to Sarvam, the LLM will focus on advanced reasoning, voice-based tasks, and fluency in Indian languages, making it uniquely suited for India's diverse population.

Sarvam AI plans to develop three key variants of its LLM:
- Sarvam-Large: Designed for advanced reasoning and complex generation tasks.
- Sarvam-Small: A lightweight model optimized for real-time interactive applications.
- Sarvam-Edge: A compact model tailored for on-device processing, enabling AI capabilities on mobile and IoT devices.
These variants aim to cater to a wide range of applications, from citizen services to enterprise solutions, ensuring adaptability across various use cases.

This initiative goes beyond technological advancement; it is a strategic move to establish critical national AI infrastructure.
The company emphasized that the goal is to create multi-modal, multi-scale foundation models that are not just functional but deeply integrated with Indian languages and societal needs.
For citizens, this means AI systems that feel familiar and culturally relevant.
For enterprises, it unlocks the potential to harness AI capabilities without concerns over data sovereignty, as all processes will remain within India's borders.

The IndiaAI Mission, approved by the Union Cabinet, is focused on scaling India's AI ecosystem by investing in compute capacity, skilled research talent, datasets, AI applications, and trusted AI practices.
One of its key initiatives is the IndiaAI Compute Capacity program, which aims to deploy over 10,000 GPUs to democratize access to AI resources for startups, researchers, and institutions.
To facilitate greater participation, especially by smaller companies, the government has also eased eligibility norms for accessing these resources, offering GPU services at globally competitive subsidized rates.
Sarvam’s selection to develop the first sovereign AI model exemplifies the mission’s objective of nurturing homegrown champions capable of competing on the global stage.

While the opportunity is historic, building a population-scale LLM is a complex challenge.
It demands seamless integration of vast datasets, engineering innovations to handle diverse languages and dialects, and fine-tuning for cultural and contextual understanding.
Additionally, unlike some global LLMs that are open-sourced, Sarvam’s model is expected to be closely managed and fine-tuned specifically for Indian use cases.
This positions it as a secure and specialized alternative in an era where data privacy and localized solutions are paramount.

Sarvam’s success could unlock a universe of possibilities, from enabling AI-driven citizen services in rural areas to building enterprise-grade AI applications with localized intelligence.
It sets the foundation for India to not merely consume global AI solutions but to become a co-creator and leader in AI innovation.
With investments from prominent venture capitalists like Lightspeed Venture Partners and Peak XV Partners, Sarvam is well-resourced to deliver on this ambitious national mission.

Q1. What is Sarvam AI’s project about?

Ans. Sarvam AI has been selected to build India’s first indigenous large language model under the IndiaAI Mission.

Q2. What are the main features of Sarvam’s AI model?

Ans. The model focuses on reasoning, supports voice tasks, and is fluent in multiple Indian languages.

Q3. How is the government supporting Sarvam AI?

Ans. Sarvam will get access to 4,000 GPUs and dedicated infrastructure to develop the model from scratch.

Q4. What are the different versions of the Sarvam model?

Ans. Sarvam is developing Sarvam-Large, Sarvam-Small, and Sarvam-Edge for different use cases.

Q5. What is the strategic importance of this project?

Ans. It enhances India’s AI sovereignty, promotes domestic innovation, and ensures secure, localized AI solutions.

Source : IE | TH