India launches three homegrown AI models, EKA-1, Dhir 2B and Dhir 5B. Two can run offline on smartphones and laptops without internet access.
India unveils three homegrown AI models, including Dhir 2B and Dhir 5B, which can operate directly on devices without an internet connection.
India has launched three homegrown artificial intelligence models, including two that can run directly on laptops and smartphones without an internet connection, as the country steps up efforts to build its domestic AI capabilities.
Backed by the IndiaAI Mission, the models are designed for Indian languages and local business requirements, with applications spanning government, education, research and private industry.
Named EKA-1, Dhir 2B and Dhir 5B, the models were unveiled at an event in New Delhi attended by government officials, including IT Ministry Secretary S. Krishnan.
Also Read: MeitY Releases Two More Indigenous AI Models Under IndiaAI Mission: What Are Their Features?
EKA-1 Targets Advanced Reasoning and Enterprise Applications
Developed by Bengaluru-based startup Soket Labs, EKA-1 is the largest of the three models. It is an open, sparse hybrid language model with 26.3 billion parameters, trained on 7.5 trillion tokens of text data.
The model is designed to handle demanding tasks such as advanced mathematics, software programming and logical reasoning. It can also function as an AI agent, carrying out a sequence of tasks independently.
EKA-1 supports English and all 22 scheduled Indian languages. Its capabilities are intended for sectors such as banking, financial services and cybersecurity, where complex information processing is important.
Also Read: IndiaAI Mission: Government Backs 20 Indigenous AI Models to Strengthen India’s Sovereign AI Push
Dhir 2B and Dhir 5B Can Work Without Internet
Unlike EKA-1, the Dhir 2B and Dhir 5B models, developed by ZenteiQ AiTech Innovations, are smaller open-source systems designed for everyday tasks.
They support English and 10 Indian languages and can search documents, extract relevant information and automate routine work. They did not disclose exact parameter counts at launch.
Their key feature is offline operation. Both models can be downloaded and run directly on smartphones and laptops without Wi-Fi or cloud servers.
Since information stays on the device rather than being sent to external servers, the models offer an additional layer of data privacy. This could make them useful for defence establishments, government departments and educational institutions handling sensitive information.
Also Read: IndiaAI Mission Launches Free AI Course for Youth and Citizens Across India
IndiaAI Mission Supports Homegrown AI Development
Developing large AI models requires substantial computing power and investment. The government’s IndiaAI Mission has supported the startups by providing subsidised access to computing infrastructure and funding research costs.
The launch highlights India’s effort to develop AI systems suited to its languages, business needs and security requirements.
The next important test for these homegrown AI models will be their adoption across institutions and businesses. Their ability to deliver practical results, particularly in offline environments where data privacy matters, will determine their wider relevance.
Samarjit Kaur is a journalist and communications professional covering technology & emerging digital trends. With a focus on clarity and context, she reports on developments shaping industries and governance. When not reporting, she chooses to plug-in and relax on her playlists and plan her next bucket-list trips!