Bodhan AI launches four multilingual AI models for Indian languages

Bodhan AI, an initiative incubated at IIT Madras, has launched four new open-weight AI models designed to support linguistic diversity in India. Developed in partnership with AI4Bharat, these models are intended to serve as Digital Public Goods to assist researchers, startups, and educational institutions.

The technological suite includes automatic speech recognition (ASR) supporting 27 languages, optical character recognition (OCR) and text-to-speech (TTS) supporting 23 languages, and machine translation covering 22 languages. These models are part of the Bharat EduAI Stack, a sovereign digital infrastructure aimed at reducing technological duplication and fostering local innovation.

In addition to the core models, the launch introduces specialized tools: a Student TutorBot for grades 6-12 that supports text and voice in 22 languages, and a Teacher Assistant Bot to help educators create lesson plans and assessments. The models were trained and optimized using NVIDIA Nemotron open models.