NEW DELHI, November 4: The IndiaAI Mission has released two sets of home-grown artificial intelligence models: EKA-1, built by Soket Labs Technology and Research, which covers all 22 scheduled Indian languages and English, and the open-source Dhīr 2B and Dhīr 5B assistants from ZenteiQ AiTech Innovations, which can run offline on local machines and phones. The models were unveiled in New Delhi on Tuesday, November 3, in the presence of S. Krishnan, Secretary of the Ministry of Electronics and Information Technology (MeitY).
Key highlights
- EKA-1 (Soket): 26.3 billion parameters, trained on about 7.5 trillion tokens; an open, sparse hybrid language model for maths, code, reasoning and agentic use.
- Languages: EKA-1 covers all 22 scheduled Indian languages plus English; Dhīr supports English and 10 Indian languages.
- Dhīr 2B and 5B (ZenteiQ): open-source multilingual assistants for secure, on-premise and offline use in defence, government, industry and education.
- Support: both teams were selected under the IndiaAI Innovation Centre (foundation models) pillar and used subsidised national AI computing capacity.
- Access: the government said such models can be made available to researchers, students and the public through its AIKosh platform.
| Model | Developer | Key features |
|---|---|---|
| EKA-1 | Soket Labs Technology and Research | 26.3 billion parameters; ~7.5 trillion training tokens; 22 scheduled languages + English; maths, code, reasoning, agentic use |
| Dhīr 2B, Dhīr 5B | ZenteiQ AiTech Innovations | Open source; English + 10 Indian languages; on-premise and offline; can run on phones; GPU and TPU support |
What the officials and developers said
Krishnan said indigenous models strengthen “India’s technological and strategic autonomy in Artificial Intelligence”. Soket’s chief executive Abhishek Upperwal said EKA-1 was tailored to India’s banking, financial services and cybersecurity landscape. ZenteiQ’s chief executive Prof. Sashikumar said the Dhīr models’ lightweight design allows them to run on mobile phones without cloud connectivity, keeping data on the device.
About the IndiaAI Mission
The Union Cabinet approved the IndiaAI Mission on March 7, 2024. It has seven pillars: compute capacity, the innovation centre, the AIKosh datasets platform, application development, FutureSkills, startup financing, and safe and trusted AI.
What it means
Models trained for Indian languages, and small enough to run without an internet connection, matter in a country where many users speak languages that global AI models handle poorly and where government and defence offices handle sensitive data. Their real test will be how widely they are used, how they perform against independent benchmarks, and whether the open-source releases attract developers.
What to watch
- Availability of the models on AIKosh and their licence terms.
- Releases from other teams selected under the IndiaAI foundation-model pillar.
S. Krishnan, Secretary, Ministry of Electronics and Information Technology. Image: Ministry of Electronics and Information Technology, GODL-India, via Wikimedia Commons. Photo source · Licence
Sources
The Tech Desk covers technology, artificial intelligence and cybersecurity. Stories are checked against at least two credible sources before publication.
