Krutrim, the AI startup founded by Ola’s Bhavish Aggarwal, has made waves with the launch of DeepSeek R1 671B in India. The company is offering access to this powerful open-source AI model at an incredibly low price of Rs 1 per million tokens throughout February. This move is generating significant buzz among developers and has the potential to reshape the Indian AI landscape.
DeepSeek R1: Affordability Meets Power
Krutrim is offering access to the powerful open-source DeepSeek R1 671B for just Rs 1 per million tokens throughout February. This promotional pricing is a fraction of the cost associated with other large language models, making it significantly more accessible to Indian developers. For context, competitors like Llama-3-8B-Instruct, Google’s Gemma-27B, and Hugging Face’s M4/idefics2-8B cost significantly more per million tokens. Aggarwal’s strategy is clear: encourage widespread adoption and experimentation to foster a thriving AI ecosystem in India. While cautioning about the DeepSeek app itself, he emphasizes the safety and potential of deploying the open-source DeepSeek R1 on Indian servers, accelerating national AI progress.
Krutrim AI Cloud and Supercomputing Ambitions
DeepSeek R1 is hosted on Nvidia’s H100 GPUs via Krutrim AI Cloud, making Krutrim the first globally to deploy this model on such powerful hardware. This deployment aims to drive AI adoption across Indian businesses and local developers. Looking ahead, Krutrim plans to deploy Nvidia’s GB200 GPUs by March 2025, aiming to build “the largest supercomputer in India” by year’s end. This infrastructure will be crucial for training and managing AI models at an unprecedented scale within the country. These ambitious infrastructure plans, coupled with the open-source approach, signal Krutrim’s commitment to building a robust AI foundation in India.
Open Source and Community Building
Krutrim has announced its intention to open-source much of its work in 2024, fostering a collaborative environment to compete with global AI leaders. This commitment to open source is further reinforced by the launch of Krutrim AI Lab, a ₹2,000-crore initiative dedicated to AI research and open-source model development. With plans to increase investment to ₹10,000 crore next year, the lab aims to address challenges like data scarcity and cultural context in Indian languages, bringing together the Indian AI community. This focus on community building is essential for creating AI solutions that are relevant and beneficial to the Indian population.
Focus on India and the Future of AI
Krutrim, meaning “artificial” in Sanskrit, has achieved unicorn status quickly. Co-founded by Bhavish Aggarwal and Krishnamurthy Venugopala Tenneti, the venture is driven by a vision to empower India’s AI landscape. The availability of DeepSeek R1 at a low cost, supercomputing plans, and open-source strategy are all part of this larger vision. Aggarwal believes that local hosting of powerful models will benefit India’s AI community, fostering collaboration and ensuring that future AI developments cater to the specific linguistic and cultural nuances of Indian users. Krutrim’s initiatives represent a significant step towards democratizing access to AI and empowering India to become a global AI powerhouse.
Subscribe:
Stay connected with the latest updates, exclusive insights, and curated content by subscribing to my newsletter.
