New Delhi: While discussions around the use of DeepSeek’s artificial intelligence (AI) model continue, Ola founder Bhavish Aggarwal has announced that India can leverage the open-source version of the model for its AI development. Krutrim, the AI platform developed by Ola to address the needs of Indian consumers, has deployed DeepSeek’s latest foundation model R1 671B on Nvidia’s H100s GPUs in India.
Krutrim’s Offerings
To promote AI development in India, Krutrim is offering the foundation model to developers at a highly competitive rate of Rs. 1 per million tokens for February. In comparison, Krutrim provides access to Meta’s Llama-3-8B-Instruct for around Rs. 16.60 per million tokens, Google’s Gemma-27B for for around Rs. 66.40, and Hugging Face’s M4/idefics2-8B for for around 16.60 per million tokens.
Earlier this month, Aggarwal announced that Krutrim would open-source its AI developments from what the company has done in 2024. “Our focus is on developing AI for India, to make AI better on Indian languages, data scarcity, (in) cultural context,” he stated.
Aggarwal also emphasized that while caution should be exercised with the DeepSeek app, India could benefit from securely deploying its open-source model on Indian servers to accelerate domestic AI advancements.
Plans Ahead
Krutrim is also set to deploy Nvidia’s GB200 GPUs by March and aims to build India’s largest supercomputer by the end of the year. Aggarwal acknowledged that while India has not yet reached global AI benchmarks, significant progress has been made in just one year. By open-sourcing its models, Krutrim hopes to foster collaboration within the Indian AI community to develop a world-class AI ecosystem.
Launched in April 2023, Krutrim became a unicorn in January 2024, achieving a valuation of over $1 billion.
Discussion about this post