The Electronics and IT Ministry announced Thursday the launch of 8 new foundational models, to develop indigenous models trained on Indian languages data, under the IndiaAI mission.

“These initiatives span multilingual and domain-specific models, scientific discovery, healthcare, and industrial innovation laying the groundwork for India’s AI leadership,” the government said in a release.

The selected startups include Avatar AI which is creating specialised ‘AI Avatars’ for all, in Indian languages across agriculture, healthcare, and governance sectors, under a 70 billion parameter model; IIT Bombay Consortium – Bharat Gen which will be building an open-source 2 billion to 1 trillion parameter models with multilingual capabilities, particularly to be deployed in agriculture, finance, legal, health, and education.

The other startups include Fractal Analytics Ltd. which is developing first large reasoning model of up to 70B parameters, for STEM disciplines; Tech Mahindra Maker’s Lab which is building an 8B parameter model for Hindi dialects and an agentic AI platform, Orion, for government applications; Zenteiq; GenLoop; Intellihealth; and Shodh AI.

Sarvam AI had been chosen for developing foundational AI model for India in April. In May, the government had selected 3 more startups including Soket AI, Gnani AI, Gan AI. With the latest launches, the total number of initiatives chosen under the IndiaAI mission for building indigenous model development has reached 12.

Until April the government had received 506 proposals. The new batch of initiatives have been chosen from the same number of proposals. This means that selected initiatives account for almost 2% of the proposals the government received.

In a post on X, Union minister Ashwini Vaishnaw wrote, “AI models being developed range from SLMs with < 10 billion parameters to LLMs with 1 trillion parameters - cutting across sectors.”

Under the ₹10,000 crore IndiaAI mission, these initiatives will be provided with subsidised GPU compute power.

"We had a target of 1000 GPUs, and today we are at 38000GPUs, which is very good progress," said Vaishnaw.

Through the IndiaAI Compute Portal, users now have access to a vast pool of AI chips: over 15,000 Nvidia H100 units, along with B200s, H200s, and others. The portal also includes AMD GPUs, AWS Inferentia 2 units, Intel Gaudi chips, and more.

Additionally, Union minister Ashwini Vaishnaw also launched 30 IndiaAI Data and AI Labs pan-India, taking the total AI lab count in India to 570. The first 27 labs are in partnership with NIELIT, established across key Tier -2 and -3 cities, and three state-of-the-art labs are set up in Mokokchung, Mhow, and Mohali with Intel. These labs will deliver foundational AI and data training courses under the FutureSkills initiative of the IndiaAI Mission.

"The AI Data Labs are intentionally located across diverse regions, reflecting our approach to inclusive growth and ensuring that technology is democratised. Our target of 570 Data Labs will be spread across the country and open doors for students going forward," Vaishnaw added

In addition to these announcement, Centre also launched the IndiaAI Fellowship Program, which will enable 13,500 scholars including undergraduates, postgraduates and PhD scholars, across different discilines.

These announcements come in the wake of the country hosting its first two-day AI Summit from February 19-20, 2026.