Science & Technology · 4 Nov 2026
यह लेख हिंदी मेंIndiaAI Mission Releases EKA-1 and Dhīr Models: Indigenous Foundation Models by Soket and ZenteiQ
Useful for UPSC mock tests · SSC mock tests · Banking mock tests · Railways mock tests
In one line: On 3 November 2026 the IndiaAI Mission released Soket's EKA-1 (26.3 billion parameters, 22 scheduled languages plus English) and ZenteiQ's open-source Dhīr 2B and 5B (English plus 10 Indian languages).
Key facts
- Released on3 Nov 2026, New Delhi
- Released byIndiaAI Mission, MeitY
- EKA-1 developerSoket Labs Technology and Research
- EKA-1 size26.3 billion parameters, ~7.5 trillion training tokens
- EKA-1 languagesAll 22 scheduled Indian languages + English
- Dhīr 2B & 5B developerZenteiQ AiTech Innovations
- Dhīr languagesEnglish + 10 Indian languages; open-source, offline-capable
- Mission pillarIndiaAI Innovation Centre (Foundation Models)
Timeline
- 7 Mar 2024Union Cabinet approves the IndiaAI Mission
- 3 Nov 2026EKA-1 (Soket) and Dhīr 2B & 5B (ZenteiQ) released at a press event in New Delhi
Why it matters for your exam
- UPSCGS III: Science and technology, indigenisation of technology; IndiaAI Mission pillars, public compute and AI sovereignty
- SSCStatic-plus-current: which company built which model, number of languages, ministry (MeitY)
- BankingSoket says EKA-1 is tailored to India's banking, financial services and cybersecurity landscape
- RailwaysOne-liners: EKA-1 = Soket, 26.3B parameters; Dhīr 2B & 5B = ZenteiQ; Cabinet approved IndiaAI Mission on 7 March 2024
Key terms
- Foundation model
- A large AI model trained on broad data that can be adapted to many tasks such as answering questions, writing code or translating.
- Parameters
- The internal numerical values a model learns during training; the count is a rough measure of model size.
- Tokens
- Small chunks of text (words or word pieces) a language model reads; training data size is counted in tokens.
- On-premise / on-device deployment
- Running a model on an organisation's own servers or on a phone instead of a remote cloud, so data stays local.
- AIKosh
- The IndiaAI Datasets Platform, one of the seven pillars of the IndiaAI Mission.
What happened
On 3 November 2026, the IndiaAI Mission under the Ministry of Electronics and Information Technology (MeitY) released two sets of indigenous foundation models at a press event in New Delhi:
- EKA-1, developed by Soket Labs Technology and Research Private Limited.
- Dhīr 2B and Dhīr 5B, developed by ZenteiQ AiTech Innovations.
MeitY Secretary S. Krishnan attended the event with senior officials of MeitY and IndiaAI and the leadership teams of both companies. The government described the models as a key milestone under the mission.
Background
The Union Cabinet approved the IndiaAI Mission on 7 March 2024, guided by the vision of "Making AI in India and Making AI Work for India". According to MeitY, the strategy aims to address India-centric challenges, create economic and employment opportunities, and position India as a global leader in AI.
The mission works through seven pillars:
- IndiaAI Compute Capacity
- IndiaAI Innovation Centre
- IndiaAI Datasets Platform (AIKosh)
- IndiaAI Application Development Initiative
- IndiaAI FutureSkills
- IndiaAI Startup Financing
- Safe & Trusted AI
Soket and ZenteiQ are among the teams selected under the IndiaAI Innovation Centre (Foundation Models) pillar to build indigenous foundation AI capabilities and large language models. The ministry says access to subsidised national AI compute and support for ancillary research expenses under the mission made these models possible.
Key details
| Feature | EKA-1 | Dhīr 2B and Dhīr 5B |
|---|---|---|
| Developer | Soket Labs Technology and Research | ZenteiQ AiTech Innovations |
| Size and training | 26.3 billion parameters; about 7.5 trillion tokens | Two models (2B and 5B) |
| Languages | All 22 scheduled Indian languages and English | English and 10 Indian languages |
| Design | Open, sparse hybrid language model | Open-source, lightweight, on-premise and offline-capable |
| Intended use | Maths, code, reasoning and agentic use | Document-grounded automation, knowledge retrieval, citizen and enterprise services |
| Target sectors (as stated) | Banking, financial services, cybersecurity | Defence, government, industry, education |
EKA-1 by Soket
Soket CEO Abhishek Upperwal said the model was built with IndiaAI Mission support and is tailored to India's banking, financial services and cybersecurity landscape, with attention to sector-specific needs.
Dhīr 2B and Dhīr 5B by ZenteiQ
The Dhīr models are meant to work as secure multilingual assistants, especially in low-connectivity or sensitive settings. ZenteiQ CEO Prof. Sashikumar said they can run on different hardware, including GPUs and TPUs, and that their lightweight design allows them to run on mobile phones without cloud connectivity, keeping data on the device.
Why it matters
- Strategic autonomy: Secretary S. Krishnan said such indigenous models strengthen India's technological and strategic autonomy in AI, and can raise productivity in government, industry, research and education.
- Language inclusion: EKA-1 covers every scheduled Indian language, which matters for services delivered in regional languages.
- Public infrastructure at work: The models show the Compute and Innovation Centre pillars feeding into actual products.
- Range of models: IndiaAI COO Sudeep Shrivastava described the releases as covering both lightweight models for everyday devices and larger models for complex enterprise uses.
- Wider access: The ministry says platforms such as AIKosh can make the models available to researchers, students and innovators.
Issues and criticism
The release itself leaves some questions open, which are worth noting for balanced answers:
- It gives no benchmark results, so performance against other models cannot be judged from the release.
- It does not name the 10 Indian languages supported by Dhīr.
- The ministry says the models can be made accessible through AIKosh; the release does not give a date or terms.
- Sector claims (banking, defence and so on) are the developers' stated aims; real-world adoption is yet to be seen.
Way forward
- Publish evaluation results, including for low-resource Indian languages.
- Make the models easy to find and use on AIKosh, with clear licences.
- Pilot them in government services where offline, on-device use helps, such as low-connectivity areas.
- Link the release to the Safe & Trusted AI pillar through testing for bias and security.
Prelims pointers
- EKA-1 was developed by Soket Labs Technology and Research.
- EKA-1 has 26.3 billion parameters and was trained on about 7.5 trillion tokens.
- EKA-1 supports all 22 scheduled Indian languages and English.
- Dhīr 2B and Dhīr 5B were developed by ZenteiQ AiTech Innovations and are open-source.
- Dhīr supports English and 10 Indian languages and can run offline, on-device.
- Both were selected under the IndiaAI Innovation Centre (Foundation Models) pillar.
- The Union Cabinet approved the IndiaAI Mission on 7 March 2024; it has seven pillars.
One-liners for SSC, Banking and Railways
- Ministry behind the IndiaAI Mission: MeitY.
- EKA-1 release date: 3 November 2026, New Delhi.
- Sector focus named by Soket for EKA-1: banking, financial services and cybersecurity.
- IndiaAI Datasets Platform: AIKosh.
- Hardware Dhīr can run on, per ZenteiQ: GPUs and TPUs, and mobile phones.
Practise more with the weekly current affairs quiz and see the SSC CGL page.
Mains practice question
"Indigenous foundation models are as much about strategic autonomy as about technology. Discuss with reference to the models released under the IndiaAI Mission in November 2026." (GS III, 250 words)
Points to cover: 1. What was released: EKA-1 (Soket, 26.3B parameters, 22 scheduled languages plus English) and Dhīr 2B & 5B (ZenteiQ, open-source, offline-capable). 2. Role of the state: Innovation Centre pillar, subsidised national compute and research support. 3. Benefits: language inclusion, data staying on device, use in sensitive sectors, autonomy in AI. 4. Gaps: public benchmarks, access terms on AIKosh, adoption and safety testing.