India has made a significant move towards technological self-reliance with the launch of three indigenous artificial intelligence models. The IT Ministry, supported by the IndiaAI Mission, has introduced EKA-1, Dhir 2B and Dhir 5B, developed domestically to address India's linguistic diversity and local business requirements while reducing dependence on foreign AI technologies.
The three models were unveiled at an event in New Delhi. IT Ministry Secretary S. Krishnan and other government officials highlighted their potential to improve productivity across government departments, research institutions, educational organisations and private businesses.
A key highlight of the launch is that two of the models are compact enough to operate directly on devices without requiring an internet connection.
What does EKA-1 offer?
EKA-1, developed by Bengaluru-based startup Soket Labs, is the largest of the three AI models. It is an open, sparse hybrid language model designed for demanding tasks such as advanced mathematics, software programming, logical reasoning and autonomous AI agent operations.
The model features 26.3 billion parameters and has been trained on 7.5 trillion tokens of text data.
Its multilingual capabilities are among its most significant features. EKA-1 can understand English and all 22 scheduled Indian languages, making it suitable for a wide range of linguistic applications.
With its advanced reasoning capabilities, Soket Labs has designed the model to serve sectors such as banking, financial services and cybersecurity.
How are the Dhir models different?
Unlike EKA-1, the Dhir 2B and Dhir 5B models are designed to be smaller and more accessible for everyday applications. Developed by ZenteiQ AiTech Innovations, the two open-source models can understand English and 10 Indian languages.
Their primary focus is on practical tasks rather than computationally intensive activities such as advanced programming or mathematics.
The Dhir models are designed to help users search through documents, extract relevant information and automate repetitive tasks. Although their exact parameter counts were not explicitly confirmed during the launch, their names are likely indicative of their respective model sizes in billions of parameters.
Why is offline AI important?
One of the most notable advantages of Dhir 2B and Dhir 5B is their ability to function without an active internet connection or a large cloud computing infrastructure.
Users can download the models onto compatible smartphones and laptops and operate them locally.
Running AI directly on a device also offers significant privacy benefits. Since information is processed locally, sensitive data does not need to be transmitted to external servers.
This capability could make the Dhir models particularly useful for organisations handling confidential information, including defence establishments, government departments and educational institutions.
Offline AI also offers greater accessibility in regions with limited or unreliable internet connectivity.
Government support for indigenous AI technology
Developing and training large AI models requires substantial computational resources and financial investment. To support Indian startups competing with international technology companies, the government has provided subsidised computing infrastructure and research funding through the IndiaAI Mission.
This assistance has helped domestic developers build AI systems tailored to India's linguistic diversity, business requirements and security needs.
The launch of EKA-1, Dhir 2B and Dhir 5B represents another step in India's efforts to establish a domestic AI ecosystem and strengthen its technological independence.
