Sarvam AI to Build India's First Homegrown GenAI Platform in Six Months under IndiaAI Mission If Sarvam's effort succeeds, it will establish India not just as a consumer of AI technologies, but as a co-creator and global leader in AI innovation

In a major stride toward building indigenous artificial intelligence (AI) capabilities, Bengaluru-based startup Sarvam AI has been selected by the Government of India to develop the country's first foundational large language model (LLM) under the INR 10,000-crore IndiaAI Mission.

Backed by venture capital giants Lightspeed Venture Partners and Peak XV Partners, Sarvam has committed to delivering the platform within six months.

The announcement was made on Saturday by Union Minister for Electronics and IT, Ashwini Vaishnaw, who expressed strong confidence in Sarvam's ability to deliver a globally competitive model. "We are confident that Sarvam's models will be competitive with the best in the world," Vaishnaw said after receiving a six-month commitment from Sarvam's founders. Highlighting the model's expected capabilities, the minister added, "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."

Access to critical technologies

To support the effort, the government will allocate 4,000 high-end GPUs to Sarvam AI for six months — critical compute power needed to build and train the massive model. These GPUs will be made available through companies such as Yotta Data Services, Tata Communications, and E2E Networks, which have been separately empaneled under IndiaAI Mission to set up AI data centres across the country.

The model, being built entirely using local talent and infrastructure, will be deeply rooted in India's linguistic and cultural landscape. Sarvam's LLM is aimed at excelling in advanced reasoning, handling voice-based tasks, and achieving fluency across Indian languages, setting it apart from globally developed models.

The road to Sarvam's selection was intense and competitive. The government had initially received over 400 proposals when it invited applications to support the development of foundational AI models earlier this year. After rigorous evaluation, Sarvam was chosen from a pool of 67 shortlisted applicants.

Sarvam's 3 models

As part of its plans, Sarvam AI is developing three model variants to cater to a wide range of use cases. Sarvam-Large will be designed for advanced reasoning and complex content generation, Sarvam-Small will optimise for real-time interactive applications requiring agility and responsiveness, and Sarvam-Edge will focus on compact, on-device processing, enabling AI capabilities on mobile and IoT devices.

"This is a crucial step toward building critical national AI infrastructure," said Vivek Raghavan, co-founder of Sarvam AI. "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."

Sarvam's vision aligns with India's broader strategic goals of achieving autonomy in AI, promoting domestic innovation, and ensuring that future AI advancements are culturally relevant and accessible to all citizens. As per the Sarvam, the model will be built, deployed, and optimised entirely within India using local infrastructure, setting a strong foundation for India's leadership in global AI development.

Co-founder Pratyush Kumar explained in media that the three model variants — Sarvam-Large, Sarvam-Small, and Sarvam-Edge — are designed to unlock a wide spectrum of possibilities: "When we build foundation models at multiple scales and modalities, it opens up a universe of applications."

The development of Sarvam's LLM also comes at a critical time in the global AI landscape. Chinese startup DeepSeek has recently disrupted the space with its low-cost, open-source foundational models, highlighting the need for nations to invest heavily in indigenous AI capabilities. DeepSeek's success, achieved using relatively inferior GPUs compared to American competitors, sent ripples through the industry, even impacting Nvidia's stock.

India's move to back Sarvam shows a determination to build world-class capabilities locally without solely relying on global tech giants.

Earlier this year, the government had also selected 10 companies to collectively supply 18,693 GPUs — significantly more than the initial target of 10,000 — showcasing the country's seriousness about accelerating its AI ecosystem.

If Sarvam's effort succeeds, it could revolutionise AI adoption across India, from powering citizen services in rural areas to enabling enterprises to build localised, intelligent applications without risking data sovereignty. More importantly, it will establish India not just as a consumer of AI technologies, but as a co-creator and global leader in AI innovation.