Mumbai: Seetha Mahalaxmi Healthcare (SML), in collaboration with the IIT Bombay-led BharatGPT initiative, has introduced ‘Hanooman AI,’ a suite of Indic large language models proficient in 22 Indian languages. Supported by Mukesh Ambani’s Reliance Industries and other engineering institutions across India, the consortium is gearing up to launch its inaugural ChatGPT-style service, BharatGPT, next month.
Led by IIT Bombay and supported by seven other IITs, the BharatGPT ecosystem enjoys backing from the Department of Science and Technology, SML, and Reliance Jio.
Key points about Hanooman:
- Nature and Scope: Hanooman comprises large language models designed for 22 Indian languages, unveiled by SML in collaboration with the BharatGPT initiative led by IIT Bombay.
- Model Sizes: Ranging up to 40 billion parameters, the initial release will feature models with parameters of 1.5 billion, 7 billion, 13 billion, and 40 billion. These models will be open-sourced upon release next month.
- Language Proficiency: Currently, Hanooman can respond in 11 Indian languages, with plans to expand its capabilities to cover all 22.
- Capabilities: Hanooman boasts multimodal AI capabilities, capable of generating text-to-text, text-to-speech, text-to-video, and vice versa content.
- Potential Applications: SML is exploring partnerships with BFSI enterprises, healthcare organisations, and mobile app providers to offer model-as-a-service solutions or develop specialised models by fine-tuning the Hanooman series.
- Challenges: Building Indian Large Language Models (LLMs) faces hurdles in acquiring quality datasets in Indian languages. Improving the quality of these datasets is crucial for text, audio, and video inputs.
- Comparison: Hanooman enters the Indic AI arena alongside models like Ola’s Krutrim, SaravamAI’s OpenHathi, and IIT-Madras’s Airavata model.
Discussion about this post