IndiaAI Mission releases indigenous foundation models EKA-1, Dhīr 2B, and Dhīr 5B

The IndiaAI Mission under MeitY released three Indian foundation models on Tuesday. The models were EKA-1, Dhīr 2B and Dhīr 5B. Officials said the launch supported India’s effort to build local AI capability. IT Secretary S Krishnan attended the event with ministry and IndiaAI officials. Leaders from both firms were also present.

Representative image

Officials called the releases a milestone under the IndiaAI Mission. The work came through the IndiaAI Innovation Centre Foundation Models pillar. That programme aims to create indigenous AI and large language models. The government said subsidised national AI compute and research support helped development. It said public AI infrastructure can speed up local innovation.

IndiaAI Mission foundation models and government support

Krishnan said such indigenous models strengthen Indias technological and strategic autonomy in artificial intelligence. They also open up opportunities to improve productivity and efficiency across government, industry, research and education, he said. Krishnan added that platforms such as AIKosh can widen access. Researchers, students, innovators and the public can test and build on them.

Both Soket Labs Technology and Research and ZenteiQ AiTech Innovations were selected under the same foundation models pillar. Officials said this support linked compute access with applied research costs. They presented the launches as a practical outcome of shared infrastructure. The event included senior officials from the ministry and IndiaAI, alongside company leadership.

IndiaAI Mission foundation models EKA-1 features

EKA-1 was developed by Soket Labs Technology and Research for Indian use cases. The model has 26.3 billion parameters and used about 7.5 trillion tokens. It supports all 22 scheduled Indian languages and English. Soket describes it as an open, sparse hybrid language model. It is designed for math, code, reasoning and agentic use.

Soket said EKA-1 can handle some inferencing on local devices. This allows certain use even without mobile data or WiFi. Soket CEO Abhishek Upperwal said the model fits India’s banking, financial services and cybersecurity needs. Abhishek Upperwal said it targets sector demands. Abhishek Upperwal said it reflects India’s changing digital ecosystem.

IndiaAI Mission foundation models Dhīr 2B and Dhīr 5B deployment

ZenteiQ AiTech Innovations developed Dhīr 2B and Dhīr 5B as open-source models. They support English and 10 Indian languages. The models were presented as secure, on-premise multilingual assistants. Dhīr 2B and Dhīr 5B were aimed at offline-capable, document-grounded tasks. These include automation, knowledge retrieval and citizen or enterprise services.

According to an official release, the focus is on low-connectivity or sensitive environments. Use cases were listed across defence, government, industrial and educational sectors. ZenteiQ CEO Prof. Sashikumar said the models run on GPUs and TPUs. Prof. Sashikumar said the lightweight design allows phone deployment. Prof. Sashikumar said cloud connectivity is not required.

Officials said on-device use can keep data local, which supports privacy and security. They said offline options can also improve access in areas with weak networks. The government linked these models to broader public infrastructure plans. It also pointed to AIKosh as a route for wider availability. The releases added new tools to IndiaAI Mission efforts.

With inputs from PTI