Speaking at the launch, S. Krishnan, Secretary of the Ministry of Electronics and Information Technology, said the public release of these foundational AI models marks an important step towards building a strong and strategic AI ecosystem in the country.
He highlighted their potential applications in several areas, including cybersecurity, and said the models could significantly improve productivity and operational efficiency across different sectors.
Emphasising their importance for secure AI deployment, Krishnan noted that the models can also operate in air-gapped environments. This capability opens up new possibilities for enterprise-based Artificial Intelligence applications in the country, particularly where secure and isolated computing environments are required.
Among the newly launched models, Eka-1 is a foundational AI model featuring 26.3 billion parameters and trained on approximately 7.5 trillion tokens. It demonstrates advanced capabilities across all 22 scheduled Indian languages, alongside English.
Developed as an open, sparse hybrid language model, Eka-1 is designed to support a range of applications, including mathematics, coding, reasoning and agentic use.
The other two models, Dhir 2B and Dhir 5B, are designed to enable offline-capable, document-grounded automation and knowledge retrieval. They can also support citizen and enterprise services across defence, government, industrial and educational sectors.
Their ability to function in low-connectivity and sensitive environments makes them relevant for applications where continuous internet access may not be available or where secure deployment is a priority.
The launch of Eka-1, Dhir 2B and Dhir 5B adds to India's indigenous AI development efforts, with the models designed to address requirements ranging from multilingual AI capabilities and advanced reasoning to offline automation and secure enterprise applications.
About The Author
Comment List