Centre to Back 20 Sovereign AI Models Under IndiaAI Mission

Kalpana Kanungo - Mumbai Bureau Jul 23, 2026 2,450 Views

The Union Government is backing 20 sovereign AI models under the Rs 10,000 crore IndiaAI Mission, comprising 12 Large Language Models and 8 Small Language Models. Confirmed in Parliament by IT Minister Ashwini Vaishnaw, the initiative provides startups with over 93 lakh GPU hours to build indigenous, low-cost AI solutions.

NEW DELHI — The Union Government has identified 20 indigenous sovereign artificial intelligence (AI) foundation model proposals for financial and compute support under the national IndiaAI Mission. In a written response to the Lok Sabha on Wednesday, July 22, 2026, Union Minister for Electronics and Information Technology Ashwini Vaishnaw confirmed that the selected projects comprise 12 Large Language Models (LLMs) and 8 Small Language Models (SLMs).

The initiative, managed by the Ministry of Electronics and Information Technology (MeitY) under its Rs 10,000 crore IndiaAI Mission, is designed to reduce dependence on foreign proprietary AI platforms while building indigenous intellectual property tailored to India's multilingual and domain-specific requirements.

High-Parameter LLMs and Domain-Specific Models in Development

The 20 approved projects span a diverse range of generative AI, speech processing, and multimodal technologies. Five of these sovereign models have already been released for public or enterprise use, with the remaining projects progressing through advanced training cycles.

Key model developments highlighted in Parliament include:

Sarvam AI: Building 30-billion and 105-billion parameter foundation models.
BharatGen: Developing multilingual foundation models covering 22 official Indian languages, led by academic consortia.
Gnani.ai: Developing specialized speech-to-speech AI models for regional voice communication.
Avataar AI: Scaling "Varya," a compact 14-billion parameter distilled video generation model designed for low-cost media rendering.

In addition to broad-spectrum foundation models, the government is backing specialized Small Language Models (SLMs) focused on healthcare diagnostics, agricultural advisory services, agentic workflows, and vernacular language translation.

Project Developer	Model Architecture / Type	Primary Focus Area
Sarvam AI	30B & 105B Parameter LLMs	High-capacity generative text & reasoning
BharatGen	Multilingual Multimodal LLM	22 Indian languages & cultural context
Gnani.ai	Speech-to-Speech SLM	Vernacular voice assistants & audio processing
Avataar AI	Distilled Video Generation (Varya)	Low-cost enterprise video rendering

Compute Infrastructure: Over 93 Lakh GPU Hours Allocated

To overcome the capital-intensive nature of training foundation models, MeitY has empanelled 15 compute service providers to deliver subsidized graphics processing unit (GPU) capacity to selected startups and research institutions.

Parliament was informed that the IndiaAI Mission has approved 237 total projects and sanctioned more than 93.18 lakh GPU hours to date. The subsidized compute infrastructure allows Indian startups to train multi-billion-parameter models at a fraction of standard commercial cloud costs.

To ensure safe deployment, the government established the AI Safety Institute and issued comprehensive India AI Governance Guidelines, creating regulatory frameworks for algorithmic auditing, deepfake detection, privacy preservation, and bias mitigation.

Official Sources Section

All details regarding government allocations, model approvals, and compute infrastructure metrics are documented in the official parliamentary response presented by the Ministry of Electronics and Information Technology (MeitY) and released via the Press Information Bureau (PIB). Program guidelines can be verified on the official IndiaAI Portal.

Quote Section

According to official parliamentary statements issued by Union Minister for Electronics and IT Ashwini Vaishnaw:

"The government has identified 20 indigenous sovereign AI model proposals for support, comprising 12 LLMs and 8 SLMs under the IndiaAI Mission. The initiative builds upon India's $300 billion IT sector to democratize AI compute, datasets, and talent across the country."

Why It Matters

Sovereign AI models ensure that digital public infrastructure, enterprise tools, and consumer applications built in India remain culturally accurate, linguistically inclusive, and secure against foreign supply chain risks. By lowering compute barriers, the IndiaAI Mission enables Indian deeptech startups to compete globally on cost efficiency and localized utility.

Key Facts at a Glance

Total Models Backed: 20 sovereign AI foundation models (12 LLMs and 8 SLMs) selected for state support.
Compute Allocation: Over 93.18 lakh GPU hours sanctioned across 237 research and startup projects.
Released Models: 5 models already live, including Sarvam AI, BharatGen, and Avataar's Varya video model.
National Reach: 58 AI Centres of Excellence approved and 27 Data and AI Labs currently operational.

FAQ Section

What is a sovereign AI model?

A sovereign AI model is an artificial intelligence platform developed, trained, and hosted locally using native datasets, ensuring data privacy, local language nuance, and technological autonomy.

How many AI models is the Indian government supporting?

Under the IndiaAI Mission, the government has identified 20 sovereign model proposals, including 12 Large Language Models (LLMs) and 8 Small Language Models (SLMs).

How does the government help AI startups train these models?

MeitY provides access to subsidized national compute infrastructure, having empanelled 15 cloud providers and sanctioned over 93 lakh GPU hours for research and development.

Source: Press Information Bureau (PIB), Ministry of Electronics and Information Technology (MeitY), IndiaAI Portal