SYLLABUS

GS-3: Science and Technology – Developments and their Applications and Effects in Everyday Life; Achievements of Indians in Science and Technology; Indigenisation of Technology.

Context: The Ministry of Electronics and Information Technology (MeitY), under the IndiaAI Mission, released three indigenous foundation models—EKA-1, Dhīr 2B and Dhīr 5B—to strengthen India's domestic AI capabilities and technological autonomy.

More on the News

• EKA-1 (Soket Labs): An open, sparse hybrid model with 26.3 billion parameters, trained on about 7.5 trillion tokens.

• It supports all 22 Scheduled Indian Languages and English and is designed for mathematics, coding, reasoning, agentic applications, banking and cybersecurity.

• Dhīr 2B and Dhīr 5B (ZenteiQ): These open-source models, with 2 billion and 5 billion parameters respectively, support English and 10 Indian languages and enable document-grounded automation, knowledge retrieval and multilingual services.

• They can run on graphics processing units (GPUs), tensor processing units (TPUs) and mobile devices without cloud connectivity, including in secure, on-premise settings.

About the IndiaAI Mission 

• Launch and outlay: Approved by the Union Cabinet in March 2024, with ₹10,371.92 crore over five years.

• Vision: "Making AI in India and Making AI Work for India."

• Implementation: Implemented by IndiaAI, an independent business division under MeitY, the Mission builds an AI ecosystem through computing infrastructure, datasets, indigenous models, skills, startup support and responsible AI.

• Progress: More than 45,000 GPUs were onboarded by June 2026. 

• By August 2026, 237 projects had accessed subsidised computing, covering 93.18 lakh GPU hours. The Mission has approved 30 India-specific AI applications, supports 8,000 undergraduate, 5,000 postgraduate and 500 PhD students, and established 27 Data and AI Labs, with 543 more identified.

About Indigenous AI Foundation Models

• Foundation Models are large-scale AI systems trained on extensive datasets and adaptable to multiple tasks. 

• Large Language Models (LLMs) specialise in understanding and generating language, while multimodal models process combinations of text, speech and images.

• IndiaAI Innovation Centre (Foundation Models) Pillar

• Objective: Develop indigenous large language and multimodal models using Indian datasets, languages and contexts to strengthen sovereign capabilities and global competitiveness.
• Selection: Twelve organisations and consortia were shortlisted, including Sarvam AI, Soket AI, Gnani AI, Gan AI, Avataar AI, IIT Bombay Consortium (BharatGen), GenLoop, ZenteiQ, Intellihealth, Shodh AI, Fractal Analytics and Tech Mahindra Maker's Lab. Separately, the government identified 20 foundation-model proposals from 506 applications—12 Large Multimodal Models and 8 Small Language Models.
• Financial Support: Selected organisations receive assistance for actual computing costs, with an additional 25% of computing expenditure earmarked for ancillary expenses such as datasets and personnel.
• AIKosh: Provides access to Indian datasets and AI models; as of July 2026, it hosted over 14,000 datasets and 331 AI models, supporting researchers, startups and innovators.

• Other Indigenous AI Models

• Param 2: Developed by BharatGen, a 17-billion-parameter multilingual foundation model supporting all 22 Scheduled Indian Languages and English.
• Vachana: Developed by Gnani.ai, a voice AI and text-to-speech model.
• Sarvam AI Models: Support multilingual document processing, translation and Indic-language understanding.

Significance of Indigenous AI Foundation Models

• Strategic Autonomy: Reduces dependence on foreign AI systems and strengthens India's ability to develop and deploy models aligned with national priorities.

• Linguistic Inclusion: Indian-language models make AI services more accessible to non-English speakers and help bridge the digital divide.

• Privacy and Cybersecurity: On-device, on-premise and air-gapped deployment can keep sensitive data within devices or organisational systems, reducing exposure to external cloud networks.

• India-specific Solutions: Models trained on Indian data and contexts can address requirements in banking, cybersecurity, defence, governance, healthcare, education and industry.

• Innovation and Productivity: Shared computing, public funding and access to open models lower entry barriers for startups and researchers, encourage domestic innovation and improve efficiency across public and private services.