From Nokia Brick Phones to Frontier AI: How Sarvam is Rewriting India's AI Playbook

New Delhi, June 15: India’s plans to move toward AI self-reliance were highlighted recently at the India AI Impact Summit 2026 in New Delhi when Sarvam AI, an Indian start-up based in Bengaluru, released its latest suite of large language models (LLMs). Models included 30- and 105-billion parameter LLMs trained from scratch on trillions of tokens across multiple Indian languages and text-to-speech, speech-to-text, and document-vision models. News agency Bloomberg called it an effort to “create something locally that can reasonably compete”, and Sarvam touted their models being “more aligned with India’s languages and cultures than products like ChatGPT and Claude”.  

Scale of Ambition 

Compared to what frontier labs have built globally though, ambitions are still relatively small. TechCrunch says Sarvam’s 105B model is already able to compete with “mid-sized open-source offerings like OpenAI’s GPT- O ́SS-120B and Alibaba’s Qwen-3-Next-80B,” systems that pale in comparison to the trillion-parameter class systems that companies like Anthropic and OpenAI are reported to be training. Sarvam has made these statements themselves, recently tweeting that its 105B Indus model “is significantly smaller than the frontier models powering global consumer chat applications today.” The company has said it plans to focus on maximizing accuracy, efficiency and alignment for Indian use cases before building larger-scale foundational models.  

Govt Support 

Government backing has been substantial. Sarvam was selected as one of 12 organizations under the IndiaAI Mission’s Innovation Centre pillar, receiving financial and compute support worth roughly ₹246.72 crore to build large language and speech models for Indian languages and public services. At the summit, Sarvam also demonstrated an AI assistant running on a basic “feature phone” without internet access, working with HMD and Qualcomm to optimize performance on existing low-end processors — a reflection of its focus on serving “a billion Indians” with small, efficient models.  

Closing the Gap 

How can India close the gap further? Three areas stand out: scaling compute access through continued government-backed infrastructure investment (the summit reportedly secured pledges toward $200 billion), deepening multilingual data pipelines across India’s 22+ official languages where global labs have less depth, and building agentic and on-device capabilities suited to low-connectivity regions rather than chasing raw parameter counts. Sarvam’s edge-first, sovereignty-driven strategy suggests India’s path may differ from a head-on race with Anthropic or OpenAI — competing instead on relevance, cost, and reach within its own massive market. 

Sources:  

• https://www.pib.gov.in/PressReleasePage.aspx?PRID=2231169&reg=3&lang=1
• https://techcrunch.com/2026/02/18/indian-ai-lab-sarvams-new-models-are-a-major-bet-on-the-viability-of-open-source-ai/
• https://www.sarvam.ai/blogs/introducing-indus