Osegi
AI·1 min read

MeitY Unveils EKA-1, Dhīr 2B and Dhīr 5B AI Models Under IndiaAI

India's IT ministry has presented three homegrown foundation models, EKA-1, Dhīr 2B and Dhīr 5B, built for specialised tasks and offline, air-gapped deployment.

Illustration of an Indian government AI initiative with neural network graphics and a map of India
Image: mint

India's Ministry of Electronics and Information Technology (MeitY) has unveiled three indigenous AI foundation models: EKA-1, Dhīr 2B and Dhīr 5B. They were developed by Bengaluru-based Socket AI and ZenteiQ under the Foundation Model initiative of the IndiaAI Mission.

What the models offer

EKA-1, built by Socket AI, has 26 billion parameters and was trained on 7.5 trillion tokens. It targets mathematics, coding and reasoning, and supports more than 100 Indian and international languages as well as programming languages. It is designed to handle multi-step tasks and execute code, and it can run in on-premises, private cloud and air-gapped environments.

Dhīr 2B and Dhīr 5B, from ZenteiQ, are mixture-of-experts models trained from scratch on Google Cloud Tensor Processing Units (TPUs). They support English and 10 Indian languages and can run directly on tested mobile hardware without an internet connection.

ZenteiQ said it is releasing the model weights under the Apache 2.0 licence, together with technical documentation and evaluation tools, through AIKosh.

Why it matters

MeitY Secretary S. Krishnan said the effort would strengthen India's strategic autonomy in AI and widen the supply of models suited to domestic needs. He pointed to uses in mathematics, coding, scientific research and cybersecurity.

Krishnan put particular weight on air-gapped deployment. "These models can be deployed on air-gapped environments, which means enterprise-based artificial intelligence, which thus far hasn't really taken off, would be possible to do a lot more of that in the country using models like this," he said.

He added that smaller models can improve productivity and efficiency across applications, and that the government will keep supporting their refinement and development. The models are available through AIKosh to researchers, students, innovators, enterprises and the general public, which Krishnan described as a wider national resource.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:meityindiaai missioneka-1dhirsocket aizenteiq

More from this day

See all →

More in: AI

See all →