BharatGen is poised to be chosen for the second phase of the IndiaAI Mission, joining seven other cutting-edge companies including Fractal, Tech Mahindra, Avataar.ai, ZenteiQ.ai, Genloop, IntelliHealth, and ShodhAI, according to sources within the Department of Science and Technology (DST). The formal announcement is expected in New Delhi next week. With these additions, the total number of participants under the Mission’s foundation model initiative will rise to 12, complementing existing members Soket AI Labs, gnani.ai, Gan.AI, and Sarvam.
Fractal, one of the new entrants, has been making headlines with its recent innovations. In May, the Mumbai-based firm introduced its open-source LLM, Fathom R1 14B, claiming it surpasses o1 mini and o3 mini in mathematical reasoning and approaches o4 mini performance levels — at a remarkably low post-training cost of just $499. Fractal is also gearing up for a public listing later this year, signaling strong market confidence.
BharatGen has been a key player in India’s sovereign AI journey. Funded by the DST, the company launched its first foundational LLM, Param 1, in May 2025. The 2.9-billion-parameter bilingual model was developed entirely in-house and featured 25% Indic data, a significant leap compared to Meta’s Llama, which reportedly had just 0.01%. The consortium behind BharatGen is led by Prof. Ganesh Ramakrishnan of IIT Bombay, with participation from IIT Kanpur, IIT Mandi, IIT Madras, IIT Hyderabad, IIIT Hyderabad, and IIM Indore. The team also released 20 speech models across 19 Indian languages on AIKosha. “We have made significant technical progress, and the announcement of our models has already come from the DST secretary,” said Ramakrishnan.
The IndiaAI Mission, supported by the Ministry of Electronics and IT, continues to scale its infrastructure at a rapid pace. In May, GPU capacity under the IndiaAI Compute Pillar expanded to 34,000 with contributions from both public and private stakeholders. Union IT Minister Ashwini Vaishnaw has reiterated that the program’s goal is to help each selected team “become a global top five player” in areas such as multilingual foundation models, speech AI, and multimodal applications.
