India launches three indigenous AI models to boost domestic capabilities

A push for technological autonomy

The Ministry of Electronics and Information Technology (MeitY) recently unveiled three indigenous artificial intelligence models in New Delhi. These models, named EKA-1, Dhīr 2B, and Dhīr 5B, were released as part of the IndiaAI Mission. The goal of this initiative is to build a domestic AI ecosystem that serves the specific needs of the country's industries and public services.

The development of these models was supported by the IndiaAI Innovation Centre's Foundation Models pillar. This framework provides domestic companies with access to subsidized national AI computing infrastructure and research funding. By providing these resources, the government aims to lower the barriers for local startups and technology firms to build advanced systems.

MeitY Secretary S. Krishnan noted that these indigenous models strengthen India's technological and strategic autonomy. The move is intended to create productivity gains across government, industry, research, and education. This initiative follows the Union Cabinet's approval of the IndiaAI Mission on March 7, 2024, which includes seven pillars such as computing capacity, datasets, and startup financing.

EKA-1 and multilingual support

EKA-1 is a foundation model created by Soket Labs Technology and Research Private Limited. It features 26.3 billion parameters and was trained on approximately 7.5 trillion tokens. The model is specifically designed to handle complex tasks involving mathematics, coding, reasoning, and agentic tasks, which are actions where an AI can perform multi-step processes.

A primary focus of EKA-1 is its broad linguistic reach. It supports all 22 languages listed in the Eighth Schedule of the Indian Constitution, as well as English. This multilingual capability is intended to help expand access to digital services and research tools for a population with significant linguistic diversity.

Soket Labs has highlighted several specific applications for the EKA-1 model. It is expected to be useful in banking, financial services, and cybersecurity. In these fields, the model's specialized language and reasoning capabilities can support unique industry-specific requirements that general models might miss.

Lightweight and secure Dhīr models

ZenteiQ AiTech Innovations developed the Dhīr 2B and Dhīr 5B models as open-source assistants. These are described as lightweight architectures, meaning they require less computing power to run. This allows them to be deployed across various platforms, including GPUs, TPUs, and even mobile devices.

One of the standout features of the Dhīr models is their ability to operate in offline-capable environments. They are designed for on-premise deployment, where the AI lives on an organization's own hardware rather than in the cloud. This is particularly important for organizations that handle sensitive information or operate in areas with limited internet connectivity.

The government indicates that these models can support document-based automation and knowledge retrieval. Potential use cases for the Dhīr models include defense, government services, industrial operations, and education. They are intended to provide secure ways for citizens and enterprises to interact with automated systems without compromising data privacy.

Implementation and next steps

While the launch marks a significant milestone for India's domestic AI development, the practical impact remains to be seen. The success of these models will depend on how they perform in real-world applications and how widely they are adopted by users. They must meet the reliability and security requirements of the organizations that choose to use them over existing global systems.

The government has indicated that platforms like AIKosh can help make these resources accessible to the public. These platforms are intended to serve researchers, students, and innovators. However, the current announcement does not confirm that every model is immediately available for unrestricted public use through these platforms.

Moving forward, the focus will shift toward independent performance evaluations and practical deployment. The IndiaAI Mission continues to work on seven pillars to ensure a safe and trusted AI environment. The success of EKA-1 and Dhīr will serve as a test case for how domestic models can compete with and complement international technology.