Govt backs 20 indigenous AI models, clears 58 centres of excellence
The government has extended support to 20 indigenous AI foundation model proposals under the IndiaAI Mission, spanning Large Multimodal Models and Small Language Models
Models already launched under this support include Sarvam AI's 30-billion and 105-billion parameter models and its Indus model, Gnani.AI's speech-to-speech model, BharatGen's multilingual models, and Avataar AI's video-generation model
58 AI Centres of Excellence have been cleared to strengthen domestic AI research, skilling, and applications across states and Union Territories
The IndiaAI Mission, approved in March 2024 with an outlay of Rs 10,371.92 crore over five years, aims to build domestic capabilities across computing infrastructure, foundation models, applications, research, and responsible/safe AI
IndiaAI Mission — Structure and Pillars
The IndiaAI Mission is a comprehensive government programme to build India's sovereign AI ecosystem, approved by the Union Cabinet in March 2024 as a Central Sector Scheme with an outlay of Rs 10,371.92 crore over five years, implemented through the IndiaAI Independent Business Division under the Digital India Corporation.
Key Details
- Built around seven pillars: Compute Capacity, Foundation Models, AIKosh (Datasets Platform), Application Development Initiative, FutureSkills, Startup Financing, and Safe & Trusted AI
- Compute Capacity pillar aims to establish a shared, subsidised high-end GPU computing ecosystem accessible to startups, researchers, and academia rather than requiring each entity to build its own infrastructure
- AIKosh is envisaged as a unified platform for datasets, pre-trained models, and tools to reduce duplication of effort across India's AI research community
- FutureSkills focuses on building an AI talent pipeline through Data and AI Labs in Tier 2/3 cities, student fellowships, and industry-aligned curricula
The 20 backed models fall under the Foundation Models pillar, while the 58 Centres of Excellence primarily support the FutureSkills and Application Development pillars — illustrating how the Mission's seven pillars work together to build a full domestic AI stack.
Sovereign/Indigenous Foundation Models and India-Specific AI Needs
A "foundation model" is a large-scale AI model (e.g., a large language model) pre-trained on broad data that can be adapted for multiple downstream tasks; "sovereign AI" refers to a nation's capability to develop and control such models independently, particularly for languages, data, and use cases specific to its context.
Key Details
- The 20 backed proposals include 12 Large Multimodal Models and 8 Small Language Models, selected to address India-specific requirements such as low-resource Indian languages, domain-specific applications, and multimodal (text-image-speech) capability
- Notable named models: Sarvam AI's 30-billion and 105-billion parameter models and its Indus model; Gnani.AI's speech-to-speech model; BharatGen's multilingual foundation models; Avataar AI's video-generation model
- Rationale for indigenous models: global foundation models are often under-optimised for Indian languages and low-resource contexts, and dependence on foreign models raises data-sovereignty and strategic-autonomy concerns
- Compute support for these efforts includes over 93 lakh GPU hours sanctioned and thousands of empanelled GPUs made available at subsidised rates to approved projects
The specific models named (Sarvam AI, BharatGen, Gnani.AI, Avataar AI) are concrete, testable examples of India's sovereign AI push — useful for distinguishing indigenous foundation-model efforts from India's use of foreign AI models.
AI Centres of Excellence and the FutureSkills Ecosystem
Centres of Excellence (CoEs) under the IndiaAI Mission are dedicated institutional hubs set up to advance applied AI research, skilling, and sector-specific solution development, distributed across multiple states to build a geographically distributed AI innovation base.
Key Details
- 58 AI Centres of Excellence have been cleared under the Mission, spread across states and Union Territories, alongside a network of India Data and AI Labs set up for skilling and applied research
- CoEs are intended to focus on priority sectors where AI can drive large-scale socio-economic impact — commonly cited focus areas include healthcare, agriculture, education, and governance
- These sit alongside 237 projects that have received subsidised compute support and 15 empanelled cloud/compute service providers under the Mission's compute-access framework
- The Mission also supports 27 India Data and AI Labs aimed at building AI skilling infrastructure outside major metro hubs
The 58 Centres of Excellence, combined with the 20 backed foundation models, represent the twin pillars (talent/application infrastructure plus model development) through which the IndiaAI Mission is operationalising India's push for AI self-reliance.
- IndiaAI Mission: approved by the Union Cabinet in March 2024, outlay of Rs 10,371.92 crore over five years
- 20 indigenous AI foundation models backed: 12 Large Multimodal Models + 8 Small Language Models
- 58 AI Centres of Excellence cleared under the Mission
- Named models: Sarvam AI (30-billion and 105-billion parameter models, Indus), Gnani.AI (speech-to-speech), BharatGen (multilingual models), Avataar AI (video generation)
- Compute support: over 93 lakh GPU hours sanctioned, 237 projects supported, 15 empanelled compute service providers
- IndiaAI Mission's seven pillars: Compute Capacity, Foundation Models, AIKosh, Application Development Initiative, FutureSkills, Startup Financing, Safe & Trusted AI