In a significant stride toward AI independence, the Indian government has selected Bengaluru-based startup Sarvam to develop the nation’s first indigenous Large Language Model (LLM). This move is seen as India's strategic response to China's rapidly emerging, cost-effective DeepSeek model.
Chosen from a pool of 67 startups, Sarvam will receive substantial support from the government, including critical computing resources, under the ₹10,370 crore IndiaAI mission. With this development, Sarvam becomes the first startup to benefit from the initiative's Standard Operating Procedures (SOPs). Government officials have confirmed that Sarvam will be granted access to 4,000 Graphics Processing Units (GPUs) for a period of six months to build and train their model efficiently.
Unlike many open-source models, Sarvam’s LLM will be a proprietary system, specifically fine-tuned for Indian languages. It aims to address India's linguistic diversity while supporting advanced reasoning, voice-based interactions, and seamless communication in multiple local languages. Government-selected companies that are establishing AI data centers across India will provide Sarvam with the required GPU resources.
According to IT Minister Ashwini Vaishnaw, Sarvam’s model will feature a staggering 70 billion parameters, positioning it on par with some of the most powerful AI systems globally. The model is designed to incorporate innovative advances in both programming and engineering, ensuring it can match the capabilities of the world’s leading AI platforms.
Sarvam plans to roll out three distinct variants to cater to different needs:
Sarvam-Large: Tailored for advanced reasoning and complex content generation tasks.
Sarvam-Small: Optimized for real-time, interactive applications.
Sarvam-Edge: Designed for lightweight, device-based operations.
The company’s approach focuses heavily on using local infrastructure and tapping into emerging Indian talent. The entire development, deployment, and optimization process will take place within India, contributing to the broader goal of strengthening the country’s AI ecosystem.
This announcement comes at a time when China's DeepSeek model is gaining considerable traction globally. Known for its affordability, accuracy, and efficiency on weaker GPUs, DeepSeek has created ripples across the AI industry. Its performance has even impacted tech giants like Nvidia, whose stock prices saw a dip following DeepSeek’s successful launch. Unlike U.S.-based models from companies like OpenAI, DeepSeek’s architecture allows it to deliver impressive results without the need for high-end computing hardware.
Given these developments, India’s move to back a powerful homegrown LLM could not be more timely. It reflects the government’s ambition to make India a leading force in the AI revolution, capable of competing with global giants on innovation, accessibility, and technological sovereignty.
Speaking about this monumental opportunity, Sarvam’s co-founder Vivek Raghavan emphasized the importance of building a strong national AI infrastructure. He stated, "Our goal is to create a fully homegrown, multi-modal, multi-scale foundation model. As we achieve this, it will pave the way for countless new applications and innovations."
Sarvam’s initiative is not just about developing another AI model; it’s about fostering a culture of indigenous innovation, nurturing homegrown talent, and preparing India for leadership in the AI-powered future.
As the project moves forward, all eyes will be on Sarvam to see how it shapes the next chapter of India's digital transformation story.
