India unveiled three locally developed artificial intelligence models Tuesday, including two designed to work offline, as the government expands domestic AI capacity for public and industrial uses.
The Ministry of Electronics and Information Technology (MeitY) announced the models under its IndiaAI Mission. Soket Labs Technology and Research Private Limited developed EKA-1, while ZenteiQ AiTech Innovations developed Dhīr 2B and Dhīr 5B.
EKA-1 is the largest of the three, with 26.3 billion parameters and about 7.5 trillion training tokens, MeitY said. The ministry says it supports English and all 22 scheduled Indian languages and is designed for mathematics, coding, reasoning and AI-agent tasks. Soket’s website directs prospective users to an early-access waitlist.
The two Dhīr models are intended for document work and information retrieval in defense, government, industry and education, according to MeitY. The ministry says they support English and 10 Indian languages.
ZenteiQ lists Dhīr 2B at 2.03 billion parameters and Dhīr 5B MoE at 5.09 billion total parameters, of which 1.79 billion are active per token. The company says the models can run offline on laptops, embedded boards and edge devices. Its website links to AIKosh downloads and lists an Apache 2.0 license.
ZenteiQ CEO Prof. Sashikumar said the models’ lightweight architecture enables on-device deployment on mobile phones. ANI reported that the company was developing free Android and iOS apps.
MeitY said both development teams received subsidized access to national AI computing infrastructure and support for ancillary research expenses. India’s Cabinet approved the broader IndiaAI Mission on March 7, 2024, with a total outlay of ₹10,371.92 crore.
ANI also reported that ZenteiQ planned a 9-billion-parameter model for December.
