India has taken another step towards strengthening its domestic artificial intelligence ecosystem with the release of three indigenous AI foundation models under the IndiaAI Mission. The models, named EKA-1, Dhīr 2B and Dhīr 5B, were developed by Indian technology companies Soket Labs Technology and Research and ZenteiQ AiTech Innovations.
The Ministry of Electronics and Information Technology (MeitY) announced the release on November 3, 2026, during an event in New Delhi attended by MeitY Secretary S. Krishnan, senior government officials and representatives of the companies involved.
The announcement represents an important development in India’s effort to build AI technologies that address domestic requirements, particularly in areas such as Indian-language computing, government services, banking, cybersecurity and enterprise applications.
EKA-1 Brings Multilingual AI Capabilities
One of the three newly released models is EKA-1, developed by Soket Labs Technology and Research.
According to the government’s announcement, EKA-1 contains 26.3 billion parameters and was trained on approximately 7.5 trillion tokens. The model supports all 22 scheduled Indian languages alongside English.
Its developers describe it as an open, sparse hybrid language model designed for several advanced applications, including mathematics, programming, reasoning and agentic AI tasks.
The multilingual capability is particularly relevant for India, where people communicate through numerous regional languages and digital services increasingly need to serve users beyond English-speaking audiences.
AI models that can understand and process Indian languages could support applications in education, public administration, financial services and customer support.
Soket Labs has also highlighted banking, financial services and cybersecurity as important areas for EKA-1. The company says the model has been developed with India’s digital ecosystem and sector-specific requirements in mind.
However, its real-world performance across these applications will need to be assessed through independent testing and practical deployments.
Dhīr 2B and Dhīr 5B Focus on Local AI Deployment
The other two models, Dhīr 2B and Dhīr 5B, have been developed by ZenteiQ AiTech Innovations.
Unlike models designed primarily for large cloud-based deployments, these models focus on lightweight multilingual assistance and deployment in environments where data privacy, connectivity and infrastructure control are important.
According to MeitY, the models support English and 10 Indian languages. They are open-source and designed for secure, on-premise deployment.
Their intended applications include document-grounded automation, information retrieval and citizen or enterprise services.
The government has identified defence, government, industrial and educational environments as potential areas of application.
An important feature highlighted by ZenteiQ is the possibility of offline-capable deployment. The company says its models can operate across different computing platforms, including GPUs and TPUs, and their lightweight architecture supports on-device applications.
This approach could be useful for organisations that cannot send confidential documents or sensitive operational information to external cloud services.
For example, an organisation could potentially deploy a local AI assistant to search internal documents, retrieve information and help employees complete administrative tasks without depending on continuous internet connectivity.
Actual hardware requirements, performance and language accuracy will still depend on the deployment configuration.
IndiaAI Mission Supports Domestic AI Development
The three models were developed under the IndiaAI Innovation Centre’s foundation-model initiative.
The IndiaAI Mission provides selected teams with access to subsidised national computing infrastructure and support for associated research expenses.
This is important because developing large AI models requires significant computing resources, specialised expertise and extensive training data.
Access to national compute infrastructure can help domestic companies experiment with foundation models without independently bearing the entire cost of building large-scale computing facilities.
The government also highlighted AIKosh as a platform through which researchers, students, innovators and other users may access AI resources.
The broader IndiaAI Mission was approved by the Union Cabinet on March 7, 2024. It operates through seven pillars covering computing infrastructure, foundation models, datasets, application development, skills, startup financing and safe and trusted AI.
Why Indigenous AI Models Matter for India
The development of domestic AI models has implications beyond creating alternatives to internationally developed products.
India has specific requirements involving language diversity, public-sector services, local business practices and data governance.
A model designed around these requirements could offer advantages in specialised applications where general-purpose international models may not always provide the desired language support or domain knowledge.
Local deployment is another important consideration.
Government departments, financial institutions, healthcare organisations and industrial companies may prefer systems that allow them to maintain greater control over their information and computing environments.
Smaller models may also make AI adoption more accessible to organisations that cannot afford expensive cloud infrastructure.
Nevertheless, developing a foundation model is only one part of establishing a successful AI ecosystem.
The next challenge will involve independent benchmarking, reliable documentation, developer access, practical integrations and long-term maintenance.
Organisations will need to assess accuracy, inference speed, hardware requirements, security and performance across different Indian languages before adopting these models for important operations.
What Comes Next?
The release of EKA-1, Dhīr 2B and Dhīr 5B adds three domestic models to India’s developing AI ecosystem.
Their long-term significance will depend on how successfully developers, businesses, educational institutions and public-sector organisations integrate them into practical applications.
For India, the opportunity is not simply to produce another collection of AI models. It is to develop technology that can address local requirements, expand access to digital services and strengthen domestic AI capabilities.
The latest announcement provides another foundation for that effort, while real-world testing and adoption will determine its eventual impact.