Krutrim, the AI startup co-founded by Ola’s Bhavish Aggarwal, has officially joined the open-source artificial intelligence race with the launch of its innovative language models. The company aims to transform AI development in India by prioritizing indigenous language processing and building essential infrastructure
. A Landmark in Indian AI Development Krutrim has introduced its first large-scale language model, aptly named ‘Krutrim,’ which means ‘artificial’ in Sanskrit. Trained on an impressive two trillion tokens, this model is crafted to understand and generate text in various Indian languages, making it one of the most comprehensive AI models addressing India’s linguistic diversity.
The company reports that Krutrim supports comprehension in 22 Indian languages and can generate content in at least 10, including Hindi, Bengali, Tamil, Kannada, Telugu, Odia, Gujarati, and Malayalam. This marks a significant advancement for AI adoption in India, where access to regional languages has been a persistent challenge.
Outperforming Global AI Models One of Krutrim’s notable claims is its enhanced performance in processing Indic languages compared to top models like OpenAI’s GPT-4. The company asserts that Krutrim utilizes nearly 20 times more Indic tokens than current AI models, providing it with a distinct advantage in accurately interpreting and responding to local dialects and linguistic subtleties. Krutrim is not limiting itself to language models.
The company has also announced the upcoming ‘Krutrim Pro,’ a more sophisticated multimodal AI model aimed at tackling complex problems and executing tasks in real-time. Krutrim Pro is set to launch in the next quarter and will feature new capabilities, including improved reasoning and interactive AI functionalities.
