Through the launch of three indigenous foundation models as part of the IndiaAI Mission, India is taking the next step towards developing its own Artificial Intelligence (AI) ecosystem. The foundation models have been developed by Indian companies Soket Labs and ZenteiQ and are designed to facilitate secure AI applications and boost productivity while reducing dependency on foreign AI technology.
Indigenous AI Models Target Secure and Enterprise Applications
A series of new models have now been launched by Soket Labs and ZenteiQ — Eka-1, Dhīr 2B and Dhīr 5B. Speaking at the launch, Electronics and IT Ministry Secretary, S. Krishnan, said that these models being made available to the public is an “important milestone” in developing India’s AI capabilities. The models can be used for various applications, including cybersecurity, and can improve productivity. They can also be used in air-gapped systems, which could make them useful for government departments, defence organisations and in the private sector for handling sensitive information.
Eka-1 is a foundational model, with 26.3 billion parameters, and has been trained using around 7.5 trillion tokens. It covers all 22 scheduled languages in India, as well as English, and can be used for tasks requiring maths, coding, reasoning and AI agents. Dhīr 2B and Dhīr 5B are designed for document automation, information retrieval and knowledge-based services. This means they can be used in an offline capacity. Potential uses for these models include government, education, defence and industry.
This demonstrates India’s increasing efforts to develop its own AI models to meet the needs of local languages, data security and enterprises. Developments like this will help improve the local AI ecosystem in India through the IndiaAI Mission.
Check this out for more information.
And for related content on India’s another AI mission with BharatGen, Click Here.
