Bhavish Aggarwal injects Rs 2K cr into Krutrim, open-sources its AI

Ola founder Bhavish Aggarwal on Tuesday announced an investment of Rs 2,000 crore in his artificial intelligence (AI) firm, Krutrim, with a commitment to invest an additional Rs 10,000 crore by next year.

Furthermore, he announced the launch of KritrimAI lab and said that India’s first AI unicorn has released its work to the open-source community, while publishing several technical reports. This move follows DeepSeek’s recent decision to open-source its own generative AI (GenAI) model.

With this latest investment, Krutrim has raised close to $280 million.

This includes the Rs 2,000 crore (approximately $230 million), and the $50 million secured from Matrix Partners India in early 2024. According to sources, the latest funds were raised through a mix of equity and debt. The company declined to comment on the specifics of the fundraising.

In India's GenAI ecosystem, Sarvam AI is the only other player to have secured funding, raising $41 million in 2023 in a round led by Lightspeed, with participation from Peak XV Partners and Khosla Ventures. Hanooman, another firm developing small language models (SLMs), has also been exploring funding opportunities.

In a X post, Aggarwal said: “We’ve been working on AI for a year; today we’re releasing our work to the open source community and also publishing a bunch of technical reports. Our focus is on developing AI for India -- to make AI better on Indian languages, data scarcity, cultural context, etc.”

He further noted that Krutrim has deployed India’s first GB200 system in partnership with Nvidia, which will be operational by March. “…and we will make it the largest supercomputer in India by end of year,” stated the founder of Ola Electric and Ola Mobility.

In a separate post on January 31, Aggarwal noted that Krutrim is utilising the DeepSeek model. “India can’t be left behind in AI. @Krutrim has accelerated efforts to develop world-class AI… Our cloud now has DeepSeek models live, hosted on Indian servers. Pricing lowest in the world,” he wrote. He added that thousands of developers have already accessed DeepSeek models hosted on Krutrim’s cloud.

Tuesday’s announcement of the AI lab is a milestone for Krutrim, marking the first time the company has shared its models with the open-source community. Until now, Krutrim had not disclosed details about its GenAI models.

Among the models released to the open-source community are Krutrim 2 and Krutrim 1, both large language models (LLMs). “While Krutrim 1 (India’s first LLM) was launched on January 24, it was a basic 7B model. We’re launching Krutrim 2 today as a much improved model,” said Aggarwal. The company claims that Krutrim 2 is a best-in-class LLM for Indic languages.

“Krutrim-2 is a 12 billion parameters dense transformer model, built on the Mistral-NeMo architecture. Our team ensured that Krutrim-2 received comprehensive training on a rich dataset encompassing English, Indic languages (hundreds of billions of tokens), code snippets, mathematical concepts, literary works, and high quality synthetically generated content. It is natively multilingual (English and 22 Indian languages) and supports a context window of 128K tokens,” said the firm’s blog.

Krutrim also introduced Chitrarth 1, a vision-language model built on Krutrim 1, capable of interpreting images and documents. Additionally, the firm unveiled Dhwani 1, a speech-language model designed for tasks such as speech translation, and announced Vyakhyarth 1, an art Indic Embedding model for use cases like search and retrieval-augmented generation (RAG).

Aggarwal further announced the development of benchmarks to assess Indian LLM performance. “Since there was no global benchmark for Indic performance, we’ve developed ‘BharatBench,’” he said. This evaluation tool is designed to assess AI models’ effectiveness in Indian languages and cultural contexts.

Whether BharatBench will gain acceptance among other Indian AI developers remains to be seen. Krutrim argued that existing benchmarks in English and Chinese, such as MMLU, ARC, and TruthfulQA, fail to capture India's linguistic and cultural nuances. "There is a specific need for a benchmark that addresses the diversity of India,” explained the startup.