In a decisive push toward AI self-reliance, the Government of India has onboarded three new startups to develop large-scale foundation models under the flagship IndiaAI Mission. The announcement, made by Union IT Minister Ashwini Vaishnaw at the “IndiaAI: Make AI in India, Make AI Work for India” event in New Delhi, marks a critical step in scaling up the country’s AI innovation ecosystem.
The latest selections Soket AI, Gnani AI, and Gan AI will each develop indigenous models targeting core national priorities such as multilingual processing, voice-based applications, and advanced text-to-speech systems.
IndiaAI Mission: Who’s Building What?
- Soket AI will lead the charge on building a 120-billion-parameter open-source model trained for India’s linguistic diversity, with use-cases in defense, healthcare, and education.
- Gnani AI is developing a 14-billion-parameter multilingual voice model for real-time speech interpretation and advanced reasoning, potentially boosting citizen-government interfaces.
- Gan AI will craft a 70-billion-parameter model focused on “superhuman” TTS (text-to-speech) capabilities, challenging global benchmarks in natural-sounding speech synthesis.
These come in addition to Sarvam AI, which was earlier tasked with creating India’s sovereign LLM ecosystem. Sarvam is already developing a 120B-parameter model designed to power public service platforms such as AI4Pragati and 2047: Citizen Connect.
GPU Power Surge: 34,333 and Counting
To back these efforts with muscle, the Centre has expanded its national AI compute capacity to 34,333 GPUs, adding 15,916 more units to the existing 18,417 pool. Industry partners such as Yotta Data Services, Sify, and Cyfuture India are among the seven vendors selected to supply this high-performance infrastructure.
This centralized GPU pool, accessible on cloud, is expected to significantly lower barriers to training and inferencing models at scale, especially for academic institutions and startups.
AI Kosh and Open Data Push
Minister Vaishnaw emphasized the importance of “common compute” as a pillar of AI democratization. “Technology should not remain in the hands of a few,” he said, reiterating the government’s vision for equitable access. Over 367 datasets have already been uploaded to AI Kosh, India’s open repository for AI-ready datasets, with more expected as part of the Data for AI initiative.
Reverse Brain Drain and Global Ambitions
“We are building a comprehensive AI ecosystem, covering compute, data, safety, and talent, that is capable of not just serving India but also competing globally,” Vaishnaw said. He urged the newly inducted teams to aim for a Top 5 global position in their respective domains.
Cybersecurity Meets AI: Hackathon Innovations
The IndiaAI I4C CyberGuard Hackathon winners were also honored on the sidelines. These groups created artificial intelligence (AI) tools to improve the National Cyber Crime Reporting Portal’s classification of cybercrime complaints. The solutions seek to use AI to modernize India’s cybercrime response infrastructure, from interpreting handwritten FIRs to transcribing audio calls.
About IndiaAI
IndiaAI is an Independent Business Division under the Ministry of Electronics & Information Technology (MeitY). It functions as the central implementation arm of the IndiaAI Mission, focusing on tech self-reliance, ethical AI development, and inclusive digital progress.
Also Read: Builder.ai Accused of Faking Business with Indian Unicorn VerSe Innovation
