India launches three homegrown AI models, EKA-1 and Dhir 2B and 5B, under the IndiaAI Mission
India’s electronics and IT ministry has released three indigenous foundation models under the IndiaAI Mission: EKA-1, a 26.3-billion-parameter model for Indian languages, banking and cybersecurity, and the compact Dhir 2B and 5B models designed to run offline on mobile phones.
India has released three indigenous artificial intelligence foundation models under the government’s IndiaAI Mission, in a step officials present as a move towards greater self-reliance in a technology now central to government, business and everyday life.
The models, EKA-1, Dhīr 2B and Dhīr 5B, were released by the Ministry of Electronics and Information Technology (MeitY) on Tuesday under the Foundation Model initiative of the IndiaAI Mission, and were reported widely in India on Wednesday.
EKA-1 was developed by Bengaluru-based Soket Labs Technology and Research. It is a 26.3-billion-parameter model trained on about 7.5 trillion tokens, built for advanced reasoning, coding and scientific research, and tailored for Indian languages, banking and cybersecurity.
The model covers all 22 scheduled Indian languages along with English, according to the Economic Times, a breadth that matters in a country where most citizens do not use English as their first language and where public services, banking and customer support must work across many tongues.
Soket Labs plans to make EKA-1’s model weights openly available within a month, with initial access provided through a waitlist, according to reports of the launch. Open weights would let researchers, start-ups and public institutions inspect, adapt and build on the model rather than relying only on a closed service.
The other two models, Dhīr 2B and Dhīr 5B, were developed by Bengaluru-based ZenteiQ AiTech Innovations. They are compact models, with two billion and five billion parameters respectively, designed to operate directly on mobile phones without requiring an internet connection.
That offline capability is the Dhīr models’ most distinctive feature. They are intended for document-grounded automation, knowledge retrieval and citizen or enterprise services in defence, government, industrial and educational settings, particularly where connectivity is low or where sensitive work cannot leave a device.
ZenteiQ has released the Dhīr model weights under the Apache 2.0 licence, along with technical documentation and evaluation tools, through AIKosh, the government’s repository of datasets and resources, according to Zee Business and other reports. The company is also planning a nine-billion-parameter model in December for researchers and engineers working on specialised problems.
At the launch event, attended by IT Secretary S. Krishnan and senior ministry and company officials, Krishnan said the indigenous models could strengthen India’s strategic autonomy in artificial intelligence. The argument is familiar from other technology programmes: a country that depends entirely on foreign models depends on foreign providers for pricing, availability, safety choices and language coverage.
The launch is part of a wider Indian push under the IndiaAI Innovation Centre’s Foundation Models pillar, which selected teams to build homegrown large language models rather than only fine-tuning systems built elsewhere. Whether EKA-1 and the Dhīr models gain real users will depend on how they perform against established global models, how easily developers can run them, and whether government departments and Indian businesses adopt them for work in Indian languages and in offline or sensitive environments.
For now, the release marks a visible milestone: India no longer has only ambitions for sovereign AI models, but three named systems, with published sizes, language coverage and licensing plans, against which future progress can be measured.
More in Technology & AI
Photo: Tencent headquarters (file image)
Technology & AITencent releases Hy4 AI model with cybersecurity training limited to defensive uses
Chinese technology company Tencent has released Hy4, a new open-source artificial intelligence model whose cybersecurity training was deliberately limited to defensive work, in an effort to reduce the risk that it could help carry out cyberattacks.
Photo: A locked computer, illustrating cybersecurity (file image)
Technology & AICrowdStrike moves to buy Fiddler AI and Dreadnode in reported $495 million bet on AI agent security
CrowdStrike has signed agreements to acquire AI agent observability firm Fiddler AI and AI security testing startup Dreadnode, adding monitoring and pre-deployment testing to its Falcon platform alongside a new AI kill switch.
Photo: An Apple iPhone (file image)
Technology & AIApple opens Siri AI to all third-party apps in iOS 27.3 developer preview, with skills that chain tasks together
Apple has released a developer preview in the iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3 betas that lets any app plug its own actions and content into Siri AI, a major widening of what the assistant can do outside Apple apps.