India set to build its own ChatGPT: Sarvam AI, Soket AI, Gan.ai, Gnani.ai lead the charge under ₹10,000 cr MeitY mission

The Ministry of Electronics and Information Technology (MeitY) is preparing to formally back four homegrown AI startups under the ambitious ₹10,000 crore IndiaAI Mission. According to sources cited by The Economic Times, the frontrunners in the race to build India’s first indigenous large language models (LLMs) include Sarvam AI, Soket AI Labs, Gnani.ai, and Gan.ai.

This initiative marks the government’s most concrete move yet in the race to develop foundational AI models that can rival those built by global giants like OpenAI and Google, but with a unique Indic edge. The move also comes in the wake of China’s DeepSeek model making waves globally, which spurred the Indian government to fast-track its own efforts. Gen AI has been the space India has been lagging behind, although Union Minister Ashwini Vaishnaw has said that Indian models would rival its global peers soon.

Each of the selected firms has submitted ambitious proposals. Sarvam AI, co-founded by former OpenAI researcher Pratyush Kumar and former Google exec Vivek Raghavan, has already begun work on a 70-billion parameter multimodal model that aims to support both Indian languages and English. Rather than a direct cash grant, Sarvam AI is expected to receive GPU compute resources valued at ₹200 crore to power its development.

Soket AI Labs, meanwhile, is eyeing a much bigger scale. CEO Abhishek Upperwal confirmed that under the ‘EKA Project’, the startup plans to develop a 120-billion parameter open-source LLM trained on a staggering 2 trillion tokens — a scale that could rival the largest global models today. While Soket didn’t disclose the funding it has sought, Upperwal said the project spans 12 months, with smaller models expected to roll out within six.

Gnani.ai and Gan.ai are focusing on smaller, more specialised models. Gnani.ai, led by CEO Ganesh Gopalan, is developing speech-to-speech AI tools for Indic languages — a crucial step in making generative AI accessible to India’s vast non-English-speaking population. The company’s co-founder Ananth Nagaraj was also involved in the drafting of the IndiaAI policy, pointing to their deep alignment with the mission’s goals.

Gan.ai, on the other hand, is said to be working on lightweight language models aimed at specific use cases, although the startup has yet to publicly comment on its proposal.

The final selection of these startups is expected to be announced in the coming days, with Union Minister Vaishnaw likely to make the official declaration. In a prior statement, Vaishnaw had hinted at the acceleration of AI model development, saying: “Very soon we will have our own LLM.”

The selection process comes after MeitY launched a ₹1,500 crore incentive scheme in January, inviting proposals to build foundational AI models from scratch. By mid-February, 67 applications had already poured in, with another 120 following in March, underscoring the urgency and enthusiasm within India’s growing GenAI ecosystem.

To support this wave of innovation, MeitY has also partnered with ten GPU-as-a-service providers — including Jio, Tata Communications, Yotta, and CtrlS — to offer access to powerful AI compute infrastructure at globally competitive rates, reportedly below $1 per hour.

With the stage set and the first cohort nearly locked in, India’s ambition to create its own ChatGPT moment — but tailored to its linguistic diversity and massive user base — is finally within reach.