IndiaAI Mission Releases EKA-1, a 26.3B-Parameter Model Covering 22 Indian Languages, and Two Dhīr Models

The IndiaAI Mission has released three indigenous foundation models: EKA-1 from Soket AI Labs, and Dhīr 2B and Dhīr 5B from ZenteiQ AiTech Innovations. EKA-1 has 26.3 billion parameters and was trained on approximately 7.5 trillion tokens, according to Analytics India Magazine.

The models were unveiled at an event in New Delhi on November 3. S Krishnan, Secretary at the Ministry of Electronics and Information Technology (MeitY), attended along with senior MeitY and IndiaAI officials.

EKA-1: Soket AI Labs

Soket AI Labs built EKA-1 as an open, sparse hybrid language model. Key details:

• Parameters: 26.3 billion
• Training data: approximately 7.5 trillion tokens
• Languages: all 22 scheduled Indian languages and English
• Capabilities: mathematics, coding, reasoning and agentic applications
• Target sectors: the company says the model is tailored to India's banking, financial services and cybersecurity sectors

"It is designed to address sector-specific requirements, understand India's evolving digital ecosystem, and deliver solutions aligned with the country's needs and priorities," said Abhishek Upperwal, CEO of Soket AI Labs.

Dhīr 2B and Dhīr 5B: ZenteiQ

ZenteiQ introduced Dhīr 2B and Dhīr 5B as open-source multilingual assistants supporting English and 10 Indian languages. They are built for document-based automation, information retrieval, and citizen and enterprise services. The intended deployment sectors are government, defence, industrial and education.

The models are designed for on-premise and offline deployment and can operate without continuous cloud connectivity. ZenteiQ says they run across computing platforms including GPUs and TPUs, and that their lightweight architecture supports on-device deployment on mobile phones.

"This allows users to access AI capabilities locally while keeping data on their devices, enhancing privacy, security and accessibility," said Prof Sashikumar, CEO of ZenteiQ AiTech Innovations.

Program and access

Both teams were selected under the IndiaAI Innovation Centre's Foundation Models pillar. The IndiaAI Mission provided access to subsidised national AI compute infrastructure and support for research expenses.

Krishnan also highlighted AIKosh, the IndiaAI platform through which the models can be made available to researchers, students and developers for further experimentation. Pricing not yet disclosed.

Sudeep Shrivastava, COO of IndiaAI, said the models reflect progress across different categories of AI development, from lightweight systems for everyday devices to models built for complex enterprise applications.

What this means

The three releases cover two different deployment profiles. EKA-1 is the larger model, with 26.3 billion parameters, and Soket AI Labs positions it for banking, financial services and cybersecurity use. The Dhīr models are smaller and aimed at local operation, where data stays on the device and no continuous cloud connection is needed.

Language coverage differs between the two developers: EKA-1 supports all 22 scheduled Indian languages plus English, while Dhīr supports English and 10 Indian languages.

Krishnan said indigenous models would strengthen India's technological autonomy in AI and create opportunities to improve productivity across government, industry, research and education. The models were built with subsidised national compute under the IndiaAI Mission, and AIKosh is the stated route for researchers, students and developers to access them.