India Launches 3 New AI Models: EKA-1, Dhir 2B and Dhir 5B Can Work Offline

India has launched three homegrown AI models under the IndiaAI Mission. While EKA-1 focuses on advanced reasoning, Dhir 2B and Dhir 5B offer offline AI capabilities.

India Introduces Three Homegrown AI Models, Including Two That Work Offline

India has taken another important step towards strengthening its domestic artificial intelligence ecosystem with the launch of three locally developed AI models. Introduced by the Ministry of Electronics and Information Technology (MeitY), with support from the IndiaAI Mission, the models are designed to address the country's language diversity, business requirements and data security needs.

The three models, EKA-1, Dhir 2B and Dhir 5B, were unveiled at an event in New Delhi. According to officials, these homegrown technologies could support productivity across government departments, educational institutions, research organisations and private businesses.

A major highlight of the launch is that the two Dhir models can operate directly on compatible devices without requiring an internet connection.

EKA-1: An Advanced AI Model for Complex Tasks

Developed by Bengaluru-based startup Soket Labs, EKA-1 is the largest among the three newly introduced AI models. It is an open, sparse hybrid language model designed to perform demanding tasks involving mathematics, programming, logical reasoning and multi-step problem-solving.

The model comes with 26.3 billion parameters and has been trained using approximately 7.5 trillion tokens of data. Its architecture is intended to support advanced reasoning and AI agent capabilities, allowing it to handle sequences of related tasks.

Supports All 22 Scheduled Indian Languages

One of EKA-1's major features is its multilingual capability. The model supports English and all 22 scheduled Indian languages, making it relevant for India's diverse linguistic environment.

Soket Labs has also focused on applications across banking, financial services and cybersecurity. Its reasoning capabilities could help organisations handle specialised tasks, analyse information and improve operational efficiency.

Dhir 2B and Dhir 5B: Compact AI Models for Everyday Use

Unlike EKA-1, which targets complex computational requirements, Dhir 2B and Dhir 5B are designed for practical, everyday applications.

Developed by ZenteiQ AiTech Innovations, both models are open-source AI systems that focus on document processing, information extraction and routine task automation.

They support English and 10 Indian languages, allowing users to interact with information across multiple languages.

Designed for Document Analysis and Automation

The Dhir models can help users search through documents, identify relevant information and automate repetitive activities. These capabilities could prove useful for offices, educational institutions and organisations dealing with large volumes of documents.

Their comparatively compact design also makes them suitable for local deployment on compatible computing devices.

How Do the Dhir AI Models Work Without the Internet?

The offline functionality of Dhir 2B and Dhir 5B is one of the most significant aspects of the announcement.

Unlike cloud-based AI services that depend on internet connectivity and remote computing infrastructure, these models can be downloaded and operated directly on supported laptops and mobile devices.

This approach allows users to access AI-powered features even in locations with limited or unreliable internet connectivity.

Better Data Privacy and Security

Running AI models locally can provide additional control over sensitive information. Since documents and user inputs can be processed directly on the device, organisations may reduce the need to transfer confidential data to external cloud servers.

This feature could make locally operated AI particularly valuable for government departments, defence establishments, educational institutions and businesses that manage sensitive information.

However, the actual privacy protections will also depend on how the models are deployed and how devices are configured.

IndiaAI Mission Supports Indigenous AI Development

Developing advanced AI models requires substantial investment in computing infrastructure, specialised hardware, research and training data.

To address these challenges, the government has been supporting domestic AI developers through the IndiaAI Mission. The initiative provides access to subsidised computing resources and helps startups reduce the infrastructure costs associated with developing large AI systems.

Support for EKA-1, Dhir 2B and Dhir 5B reflects India's broader ambition to develop artificial intelligence technologies within the country rather than depending entirely on international platforms.

What Does the Launch Mean for India's AI Industry?

The introduction of these three models highlights India's growing focus on building an independent AI ecosystem.

While EKA-1 targets advanced reasoning and enterprise-level applications, the Dhir models focus on accessibility, offline functionality and local data processing.

Their multilingual capabilities could also help expand AI adoption among users who prefer communicating in Indian languages.

As domestic developers continue to improve these technologies, homegrown AI models could become important tools for businesses, public institutions and individual users.

Conclusion

The launch of EKA-1, Dhir 2B and Dhir 5B marks an important development in India's efforts to strengthen its indigenous AI capabilities. With advanced reasoning, multilingual support and offline functionality, these models aim to address different technology requirements. Backed by the IndiaAI Mission, the initiative could help India build a more accessible, secure and self-reliant artificial intelligence ecosystem.