India’s artificial intelligence mission is moving into its next big phase. The government has named eight organisations, including IIT Bombay, Tech Mahindra and Fractal Analytics, to build foundational large language models (LLMs) under the IndiaAI Mission. IT Minister Ashwini Vaishnaw announced the names during the AI Impact Summit 2026 in New Delhi on September 18.
The most ambitious of these projects is being led by IIT Bombay through its BharatGen consortium. The institute has been tasked with developing an LLM that will have one trillion parameters, a scale that places it among the largest AI models currently being attempted anywhere in the world. To support this effort, the IndiaAI Mission has approved financial assistance worth Rs 988.6 crore.
In the context of AI, parameters are the learned internal variables that allow a model to recognise patterns and relationships in data. A higher parameter count often translates into a more sophisticated understanding of language. For India, a trillion-parameter model is seen as a major milestone in developing home-grown AI capabilities that can compete with global leaders.
The full list of entities announced includes Avataar AI, IIT Bombay Consortium – BharatGen, Fractal Analytics Limited, Tech Mahindra Ltd, Zeinteiq Aitech Innovations, Genloop Intelligence Pvt Ltd, NeuroDX (Intellihealth) and Shodh AI. Each of these organisations will play a role in building foundational LLMs aimed at different applications.
Tech Mahindra, which has already been working on its in-house Indic model called Project Indus, welcomed the recognition. In a statement, the company said it was “proud to be recognised” as part of the IndiaAI Mission. “This announcement comes on the back of building our own and India’s own Indic LLM, Project Indus. Built completely in-house and at frugal cost, the journey of having Project Indus as open source to creating sovereign LLMs has been a learning and rewarding experience,” it added.
This latest move builds on earlier rounds of selection under the Mission. In May 2025, three startups — SoketAI, Gnani.ai, and Gan AI — were chosen to create India’s first home-grown foundational models. A month before that, four other startups, including Sarvam AI, were identified to focus on specialised AI systems. To ensure these teams have the resources needed, the government has also worked with cloud and data providers to expand access to GPUs, which are essential for training large AI models.
Minister Vaishnaw said the progress of the first batch of models has been promising. “The models that were selected earlier are progressing really well and I am confident that by the time the AI Impact Summit gets underway in February 2026, India will have a model or models ready,” he noted.
Looking ahead, the government also plans to unveil an AI framework in the coming days. The framework is being prepared jointly by the Ministry of Electronics and Information Technology (MeitY) and the Office of the Principal Scientific Adviser. It is expected to provide clear guidelines on how AI systems should be built and deployed responsibly in the
