New Delhi: India has taken another step toward self-reliance in the field of Artificial Intelligence (AI). In collaboration with the ‘IndiaAI Mission,’ the Ministry of Electronics and Information Technology (MeitY) has launched three new indigenous AI models: EKA-1, Dhir 2B, and Dhir 5B. These models have been developed with a focus on Indian languages, local requirements, and facilitating operations across various sectors. Notably, Dhir 2B and Dhir 5B can even run on mobile devices and laptops without an internet connection.
These AI models were unveiled at an event held in New Delhi. During the event, officials—including S. Krishnan, Secretary at the Ministry of Electronics and Information Technology—explained that these models could be utilized in government offices, educational institutions, research centers, and private companies. The objective is to enhance productivity while reducing dependence on foreign AI technologies.
Advanced Features of the EKA-1 Model
Among the three, EKA-1 is the largest and most advanced system. It was developed by the Bengaluru-based startup, Soket Labs. As an open, sparse hybrid language model, it is designed for complex tasks such as mathematics, software programming, and logical reasoning. Furthermore, it can function as an AI agent, capable of executing multi-step tasks.
EKA-1 consists of 26.3 billion parameters and has been trained on 7.5 trillion tokens of text data. A key feature is its ability to understand English as well as all 22 Indian languages listed in the Eighth Schedule of the Constitution. It is also tailored for use in sectors such as banking, financial services, and cybersecurity.
What are the key features of Dhir 2B and Dhir 5B? Compared to EKA-1, Dhir 2B and Dhir 5B are smaller AI models designed for everyday tasks. Developed by ZenteiQ AiTech Innovations, these open-source models are capable of understanding English as well as 10 Indian languages.
They can be used for tasks such as searching for information within documents, extracting relevant data, and automating routine processes. The ‘2B’ and ‘5B’ in their names refer to their parameter sizes; due to their compact nature, they are optimized to run on standard devices.
Can be used without an internet connection
A key feature of Dhir 2B and Dhir 5B is their ability to function offline. They can be downloaded onto mobile phones or laptops and used without an internet connection, eliminating the need for a constant link to a cloud server.
Since the AI models operate directly on the device, there is less need to transmit a user’s private data to external servers. This helps enhance data privacy and security, making these models particularly useful for government departments, the defense sector, and educational institutions.
IndiaAI Mission strengthens indigenous AI
Developing and training large AI models is highly expensive, requiring state-of-the-art computing infrastructure and immense processing power. In this context, India’s ‘IndiaAI Mission’ is providing essential resources and computing facilities to the country’s startups and research institutions.
Through this initiative, Indian companies are receiving support for AI research and model development. The launch of these three indigenous AI models is considered a significant step toward developing a robust AI ecosystem within India.
