Ola founder and chairman Bhavish Aggarwal announced the Krutrim AI lab on Tuesday. The AI-focused research lab will be the central hub for all the future model releases of Ola Krutrim (commonly referred to as Krutrim). Alongside, several new India-focused open-source artificial intelligence (AI) models were also released, with the main highlight being the Krutrim-2 AI model. Aggarwal also announced an investment of Rs. 2,000 crores in Krutrim and committed an investment of Rs. 10,000 crores by the next year.
Krutrim AI Lab and New Models Released
In a post on X (formerly known as Twitter), Aggarwal highlighted that his AI firm has been working on developing new AI innovations since the formation of the company. Showcasing Krutrim’s achievements in the last year, the Ola chairman also released several new open-source AI models. “Our focus is on developing AI for India – to make AI better in Indian languages, data scarcity, cultural context etc.,” he added.
Among the models released, the biggest highlight is the Krutrim-2, a 12-billion-parameter model which comes as the successor to the Krutrim-1 AI model. It is a dense transformer model, built on the Mistral-NeMo-12B-Instruct architecture. The natively multilingual model can generate responses in English and 22 Indian regional languages and supports a context window of 1,28,000 tokens. The model is currently available via a Hugging Face listing with the Krutrim Community License for academic and research-oriented usage.
Krutrim also announced the release of Chitrarth-1, a multilingual Vision Language Model (VLM) built on Krutrim-7B. It uses a SIGLIP vision encoder to extract and process visual information. The company stated that the AI model is trained on multilingual image and text data and can work across nine Indian regional languages, including Assamese, Bengali, Gujarati, Kannada, Malayalam, Marathi, Odia, Tamil, and Telugu. It also understands English.
For AI-based speech translation, the company released the Dhwani-1 automatic speech recognition (ASR) model. It supports translation between Indic Languages and English. The supported languages are English, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Tamil, and Telugu. It is accompanied by the Krutrim Translate text translation model.
Aggarwal also announced the release of Vyakyarth-1-Indic-Embedding, a multilingual sentence-transformer model designed for semantic textual similarity, search, clustering, and classification across 100 languages. He added that this model can be useful for cross-lingual natural language processing (NLP) applications.
In addition, Aggarwal highlighted that the company had also developed a new benchmark dubbed BharatBench that measures the performance of an AI model on indic languages. He claimed that the benchmark captures the unique linguistic and cultural nuances of India, which is not seen in other tests.
“We’re nowhere close to global benchmarks yet but have made good progress in 1 year. And by open-sourcing our models, we hope the entire Indian AI community collaborates to create a world-class Indian AI ecosystem,” Aggarwal said.
Krutrim is also partnering with Nvidia to deploy the tech giant’s Blackwell-based GB200 GPU for AI-enabled workloads. The infrastructure will be launched by March.
