Ola, an Indian AI startup, has launched Krutrim, a family of large language models (LLMs) trained on a number of different Indian languages.
Krutrim comprises two main variants: The base model, which has been trained on two trillion tokens and is capable of generating text in 10 Indian languages, and Krutrim Pro, which is expected to support text, speech, and video with more advanced problem-solving capabilities. While the base version will be launched immediately, the company plans to release the Pro model in the upcoming quarter.
Ola's Krutrim model utilizes an in-house AI chip named “Silicon,” differentiating itself from established players like NVIDIA in the AI chip market. The company claims that its Krutrim model outperforms GPT-4 in Indic language performance and surpasses Llama-2 in qualitative evaluation and reasoning capabilities.
By using this site, you agree to allow SPEEDA Edge and our partners to use cookies for analytics and personalization. Visit our privacy policy for more information about our data collection practices.