New Delhi: The government has selected three additional AI startups, Soket AI, Gnani AI, and Gan AI, to develop indigenous foundation models, following the earlier selection of Sarvam AI.
This is part of the IndiaAI Mission’s Foundation Model pillar, which seeks to build sovereign large language models (LLMs) trained on India-specific data to address national priorities in governance, inclusion and innovation.
Since the launch of this initiative, it has attracted significant interest, with 506 proposals received till April. The four selected firms are now tasked with building India’s next generation of open-source LLMs to power AI use cases across sectors.
Soket AI will develop India’s first 120 billion parameter open-source foundational model optimised for linguistic diversity and tailored for applications in defence, healthcare and education.
Meanwhile, Gnani AI will build a 14 billion parameter Voice AI foundation model designed for real-time multilingual speech processing with high-level reasoning capabilities.
Gan AI will focus on text-to-speech innovation, developing a 70 billion parameter multilingual model aimed at achieving “superhuman” performance in TTS, rivalling global leaders.
The latest additions come after Sarvam AI was awarded the mandate to build a 120 billion parameter Sovereign LLM, following its earlier launches of Sarvam-1 (2B parameters) and Sarvam-M (24B parameters), both with hybrid reasoning capabilities.
As reported, Sarvam’s new model will drive flagship initiatives like “2047: Citizen Connect” and “AI4Pragati”.
To support the growing model development efforts, the IndiaAI Mission has also ramped up the AI compute capacity significantly.
Union IT Minister Ashwini Vaishnaw announced the empanelment of 15,916 new GPUs, bringing the total to 34,333 GPUs now available via the IndiaAI Compute Platform.
These resources, accessible through a unified cloud portal, aim to make high-performance computing affordable and scalable for startups, researchers and enterprises.
Seven companies, Cyfuture India, Ishan Infotech, Locuz, Netmagic, Sify, Vensysco and Yotta, have submitted commercial offers for deploying GPU infrastructure, which includes top-end chips from Nvidia (H100, H200, B200), AMD (MI300X, MI325X) and Intel (Gaudi series).
These companies will join existing cloud partners like Jio Platforms, Tata Communications, NxtGen, CtrlS and E2E Networks to create a distributed AI infrastructure across the country.
