Quick answer: EKA-1, Dhīr 2B and Dhīr 5B are Indian-built AI foundation models released by the IndiaAI Mission of the Ministry of Electronics and Information Technology (MeitY) on 3 November 2026. EKA-1 comes from Soket Labs and covers all 22 scheduled languages plus English, while ZenteiQ’s open-source Dhīr models are designed to run offline and on-premise, including for defence users.
Why EKA-1 and Dhīr are in the news
According to a PIB release from MeitY, the IndiaAI Mission released the three models at a press event in New Delhi on 3 November 2026. MeitY Secretary S. Krishnan attended along with IndiaAI officials and the leadership of the two companies. Both firms were chosen under the IndiaAI Innovation Centre (Foundation Models) pillar of the mission, which gave them subsidised access to national AI compute and support for ancillary research costs.
Krishnan said home-grown models of this kind strengthen India’s “technological and strategic autonomy” in AI, and added that such models can be opened to researchers, students and innovators through AIKosh, the mission’s datasets platform.
Key facts: EKA-1 vs Dhīr at a glance
| Feature | EKA-1 | Dhīr 2B and Dhīr 5B |
|---|---|---|
| Developer | Soket Labs Technology and Research Pvt Ltd | ZenteiQ AiTech Innovations |
| CEO quoted | Abhishek Upperwal | Prof. Sashikumar |
| Size / training | 26.3 billion parameters; about 7.5 trillion tokens | Lightweight; can run on mobile phones |
| Languages | All 22 scheduled Indian languages and English | English and 10 Indian languages |
| Type | Open, sparse hybrid language model | Open-source multilingual assistants |
| Built for | Maths, code, reasoning and agentic tasks; tuned for banking, financial services and cybersecurity | Secure, on-premise, offline use in defence, government, industry and education |
| Hardware | Not specified in release | GPUs and TPUs; on-device without cloud |
- Releasing body: IndiaAI Mission under MeitY.
- Release date: 3 November 2026, New Delhi.
- Pillar: IndiaAI Innovation Centre (foundation models).
- Language reach: EKA-1 covers all 22 languages of the Eighth Schedule.
- Defence relevance: Dhīr is pitched for low-connectivity or sensitive environments, where data must stay on the device.
What these terms mean (for non-science students)
A foundation model is a large AI model trained on huge amounts of text that can then be adapted to many tasks, such as answering questions, summarising files or writing code. Parameters are the internal values the model learns during training, and tokens are the small pieces of text it reads. “Agentic use” means the model can carry out multi-step tasks, not just reply to one question.
The Dhīr point that matters most for the armed forces is offline, on-premise deployment. ZenteiQ’s CEO said the models can run locally on a phone, so data does not leave the device. For units working in border areas with poor networks, or handling classified documents, a model that does not depend on a foreign cloud server is a security advantage. That is the practical meaning of “strategic autonomy” in AI.
Where this fits in the IndiaAI Mission
The Union Cabinet approved the IndiaAI Mission on 7 March 2024 under the motto “Making AI in India and Making AI Work for India”. PIB lists its seven pillars as IndiaAI Compute Capacity, IndiaAI Innovation Centre, IndiaAI Datasets Platform (AIKosh), IndiaAI Application Development Initiative, IndiaAI FutureSkills, IndiaAI Startup Financing, and Safe & Trusted AI. EKA-1 and Dhīr are outputs of the second pillar and were trained using the first. For the mission’s budget, GPU numbers and platforms such as Bhashini, see our earlier explainer linked below.
IndiaAI COO Sudeep Shrivastava said the releases show progress at both ends, from light models for everyday devices to heavier enterprise systems.
Exam angle: how NDA, CDS and CAPF can ask this
- Science and technology questions in CDS and AFCAT often ask “which ministry” or “which mission” launched a new platform; here the answer is MeitY’s IndiaAI Mission.
- The Eighth Schedule (22 languages) is core polity GK for NDA and CDS; EKA-1’s language coverage is a natural hook.
- CAPF Paper II and SSB discussions often turn to technology and internal security; offline AI for defence and sensitive environments is a good talking point.
- Know the seven IndiaAI pillars and that AIKosh is the datasets platform.
Practice MCQs
Q1. EKA-1, released under the IndiaAI Mission in November 2026, was developed by:
(a) Sarvam AI (b) Soket Labs (c) ZenteiQ AiTech (d) BharatGen
Answer: (b) — Soket Labs Technology and Research developed EKA-1.
Q2. How many Indian languages do the Dhīr 2B and Dhīr 5B models support, besides English?
(a) 5 (b) 10 (c) 15 (d) 22
Answer: (b) — they support English and 10 Indian languages.
Q3. AIKosh under the IndiaAI Mission is a:
(a) GPU cluster (b) Datasets platform (c) AI safety institute (d) Start-up fund
Answer: (b) — PIB lists it as the IndiaAI Datasets Platform.
Q4. The IndiaAI Mission was approved by the Union Cabinet on:
(a) 26 January 2023 (b) 7 March 2024 (c) 15 August 2024 (d) 1 April 2025
Answer: (b) — approved on 7 March 2024.
Frequently asked questions
What is EKA-1?
EKA-1 is an Indian foundation AI model by Soket Labs with 26.3 billion parameters, trained on about 7.5 trillion tokens, and covering all 22 scheduled Indian languages and English.
Who developed Dhīr 2B and Dhīr 5B?
ZenteiQ AiTech Innovations developed them as open-source multilingual models for secure, offline and on-premise use.
When were EKA-1 and Dhīr released?
The IndiaAI Mission under MeitY released them on 3 November 2026 at an event in New Delhi.
Why are the Dhīr models useful for defence?
They can run on local hardware and even mobile phones without cloud access, so sensitive data stays on the device and the tools work in low-connectivity areas.
Under which pillar of the IndiaAI Mission were these models built?
Under the IndiaAI Innovation Centre (Foundation Models) pillar, with compute support from the mission.
Also read on NCA: Artificial Intelligence in India and the IndiaAI Mission | ISRO’s next decade | Monthly defence current affairs PDFs | Online NDA coaching
Sources: PIB (MeitY), 3 Nov 2026, Indian Television. Facts as reported up to 5 November 2026.
Featured image background (representative): 极客湾Geekerwan / Wikimedia Commons, CC BY 3.0.
