Archive · 12 Sep 2026
Bodhan AI and AI4Bharat Release Open Indic-Language Models
Bodhan AI and AI4Bharat have created new AI models for Indian languages.
These models are open-weight, which means developers can use and adapt them.
One model understands spoken language.
Another turns written text into speech.
A third translates between languages.
A fourth can read words from images and documents.
The groups used NVIDIA NeMo to train some of the models.
They say the models will help more people build AI tools for India’s many languages.
Bodhan AI and AI4Bharat announced open-weight foundational AI models for Indic languages.
The model suite covers speech recognition, text-to-speech, machine translation and optical character recognition.
The models are named Indic-Transcribe, Indic-Speak, Indic-Translate and Indic-OCR.
Bodhan AI and AI4Bharat use NVIDIA NeMo to train models for speech recognition, translation and OCR.
The organizations aim to help Indian developers deploy, adapt and fine-tune multilingual and multimodal AI systems.
- Who
- Bodhan AI, an IIT Madras Centre of Excellence in AI for Education incubated at IIT Madras, and AI4Bharat.
- What
- They announced a suite of open-weight foundational AI models for Indic languages.
- When
- September 12, 2026.
- Where
- India’s Indic-language AI ecosystem; the articles do not specify a launch location.
- Why
- To give developers and model builders open models they can deploy, adapt and fine-tune, supporting multilingual and multimodal AI innovation.
- Organizations
- Bodhan AI and AI4Bharat
- Model type
- Open-weight foundational AI models
- Languages
- Indic languages
- Capabilities
- Speech recognition, text-to-speech, machine translation and optical character recognition
- Model names
- Indic-Transcribe, Indic-Speak, Indic-Translate and Indic-OCR
- Training framework
- NVIDIA NeMo
- Publication date
- September 12, 2026