Ola's Krutrim, Lenovo Team Up to Create 700 Billion Parameter LLM Model
Together with Lenovo, Ola's artificial intelligence venture, Krutrim, is creating the 700 billion-parameter large language model (LLM) known as Krutrim 3.
To support its AI projects, the business is also constructing the biggest supercomputer in India.
This is just a month after Bhavish Aggarwal, the founder of Ola, declared that he will invest Rs 2,000 crore in his AI project, with an additional Rs 10,000 crore pledged by the following year.
Ola is also establishing a lab for AI research. It is also developing its own AI processors and cloud infrastructure in the interim.
At the Lenovo event in Mumbai, Navendu Agarwal, Chief Information Officer at Ola, announced the news, saying that Krutrim 3 will be India's response to the global AI race.
This comes after Bhavish Aggarwal, the creator of Ola, pledged to spend Rs 2,000 crore in Krutrim, with ambitions to double that amount to Rs 10,000 crore the following year.
At the moment, the company has 700 people working on AI research and development.
Since its start, Krutrim has introduced two open-source AI models: Krutrim 1, which had seven billion parameters and was released in late 2023, and Krutrim 2, which had twelve billion parameters.
To lessen dependency on other technology, the business has emphasized the necessity for India to build its own hyperscale AI infrastructure and sovereign cloud.
Along with using GB200 superchips in partnership with the massive American semiconductor company Nvidia, Krutrim is also creating its own AI chip as part of this approach. In order to improve data privacy and save training expenses, the company previously installed AI models from Chinese startup DeepSeek on its domestic servers.
Bhavish Aggarwal used social media to highlight the importance of the most recent advancements, claiming that India requires extensive AI infrastructure in order to compete globally. He also asked for ideas on the future supercomputer's name.