Government launches three homegrown AI models under IndiaAI Mission

The Ministry of Electronics and Information Technology (MeitY) has unveiled three AI models built in India under the IndiaAI Mission. The models, called EKA-1, Dhir 2B and Dhir 5B, were launched at an event in New Delhi on Wednesday, 4 November.

EKA-1 comes from Soket Labs Technology and Research. Soket says the model is an open, sparse hybrid language model with 26.3 billion parameters. It was trained on about 7.5 trillion tokens, the chunks of text an AI model learns from. The model supports all 22 scheduled Indian languages and English. Soket says it is built to handle maths, coding, reasoning and AI agents, and that it has been developed with a focus on banking, financial services and cybersecurity.

The two Dhir models come from ZenteiQ AiTech Innovations. Both are open-source and support English and 10 Indian languages. Unlike many AI tools that depend on cloud servers, they can run locally on a device and work without an active internet connection. ZenteiQ says they can assist users with finding files, looking up facts and running routine workflows.

ZenteiQ says the Dhir models could suit defence, government, education and industry, where data privacy and security matter most. Because they run on the device itself, sensitive files need not be sent to outside servers.

The launch is part of India's wider effort to build its own AI capability and rely less on foreign technology. A locally run model could mean more control over where data goes, especially in sectors that handle confidential information.

How many people or organisations are already using the models, and when they will be widely available, has not been stated. Independent testing of the models' performance has not been reported.