Tech Mahindra launches 8-billion parameter Hindi-first education LLM at AI Impact Summit

Tech Mahindra (Courtesy: Reuters photo)

Tech Mahindra on February 20 announced the launch of a new education-focused large language model (LLM) under the IndiaAI Mission at AI Impact Summit in Delhi. The Hindi-first LLM was built by advancing the IT firm’s existing Project Indus.

The company said the new model is an upgraded version of its earlier 1.2-billion-parameter foundational LLM and has now been scaled to an 8-billion-parameter architecture. Moneycontrol had earlier reported on the model's usecases in November.

The latest iteration has been purpose-built for education use cases, with a focus on democratizing access to high-quality learning in Hindi and other Indian languages.

The education LLM also supports agentic AI capabilities, allowing developers to build autonomous AI agents that can understand and respond fluently in Hindi. It is ultimately designed to enable a deeper foundational understanding of subjects such as physics and other disciplines, while also embracing India’s rich linguistic and cultural diversity.

To develop the model, Tech Mahindra used NVIDIA’s AI stack, including the NeMo framework and deployment via NVIDIA NIM microservices, to enable production-grade scalability and performance. The company also generated around half a billion synthetic tokens using NVIDIA NeMo Data Designer to address data scarcity in select Indian languages.

Nikhil Malhotra, Chief Innovation Officer and Global Head of AI and Emerging Technologies at Tech Mahindra, said global foundational models are often not tailored for countries with deep linguistic and cultural diversity like India, particularly in the education sector.

"Through Project Indus, our collaboration with NVIDIA directly addresses this gap by delivering a Hindi-first, sovereign AI model that enables scalable, relevant, and accessible AI-powered learning and citizen-centric services for India," Malhotra added.

John Fanelli, Vice President, Enterprise Software, NVIDIA, added, "The global push for sovereign AI is accelerating demand for foundation models tailored to local languages and cultural contexts. By leveraging NVIDIA AI Enterprise, Tech Mahindra delivers the production-ready performance, reliability and scale required to power Project Indus."

This is the fifth sovereign foundation model launched under IndiaAI Mission in the last three days of AI Impact Summit. Previously, Sarvam, Gnani.ai, BharatGen and Fractal Analytics had released their models spanning across segments and usecases including healthcare reasoning, AI voice model, efficiency improvement model, and multilingual LLM to enable AI adoption for governance, agriculture, enterprise.

Also read: India rolls out three sovereign AI models Sarvam AI, Gnani.ai, BharatGen to take on Big Tech