Krutrim Unleashes Kruti: India's First Agentic AI Goes Beyond Chatbots

Krutrim, an artificial intelligence venture founded by Ola's Bhavish Aggarwal, has announced the launch of Kruti, an AI assistant described as India's first "agentic" AI.[1][2] This development marks a significant step for the company, which aims to build a comprehensive AI stack, including foundational models, cloud infrastructure, and even silicon chips, tailored for the Indian market.[1][3][4] Kruti is designed to go beyond the capabilities of standard chatbots by not just responding to user prompts, but also by proactively listening, adapting, and taking action in the user's language.[1][5][6]

At its core, Kruti is powered by Krutrim's V2 Large Language Model (LLM), which is a 12 billion parameter model built on the Mistral-NeMo architecture.[7][8][9] This model has been trained on extensive datasets, including a significant corpus of Indian languages and context-specific data, encompassing over two trillion tokens.[10][11][12] The training data includes web data, books, and synthetic data to ensure robust linguistic representation and understanding of cultural nuances.[7][9] Krutrim has stated that its LLMs are trained to understand over 10 Indian languages, with plans to expand support to all 22 official languages.[13][10][11] The company claims its models, like Krutrim-2, outperform comparable, often larger, models on Indic language benchmarks and show competitive performance on English benchmarks as well.[7][9] To enhance scalability and cost-efficiency in deployment, Krutrim also leverages open-source models, and has even started hosting Meta's Llama 4 models on its domestic servers.[5][13] This strategy is intended to reduce reliance on expensive Western cloud platforms and offer more affordable AI solutions for Indian developers and enterprises.[13][4]

The introduction of Kruti as an "agentic AI" assistant signifies a move towards more autonomous and proactive AI systems.[14][6] Agentic AI is designed to understand context, make decisions, and execute actions to achieve preset objectives with limited human supervision, effectively bridging the gap between static programming and dynamic adaptability.[15][16][17] Unlike traditional chatbots that primarily react to explicit instructions, an agentic assistant like Kruti is envisioned to anticipate user needs, manage tasks like booking appointments or managing emails, and potentially integrate with other applications to perform actions such as booking a cab without requiring users to switch between apps.[5][13][14][10] This capability to perform multi-step tasks autonomously is what Krutrim believes will set Kruti apart and represent a "leap beyond chatbots".[1][5] For example, Krutrim has teased capabilities such as Kruti assisting with food-related queries, from what to eat and where to order, to how to cook it.[5] The assistant is also expected to learn from user interactions over time to provide personalized experiences.[13][18]

Krutrim's broader strategy extends beyond just AI assistants. The company is positioning itself as a full-stack AI provider for India, aiming to build a self-reliant AI ecosystem.[1][3][19] This includes the Krutrim Cloud, a platform offering AI computing infrastructure, Model-as-a-Service (MaaS), and GPU-as-a-Service to developers and enterprises.[13][11][20] Krutrim Cloud hosts both Krutrim's proprietary models and various open-source models, aiming to foster innovation within India's AI sector at lower costs.[13][10][11] The company has reported that over 25,000 developers are using Krutrim Cloud, with 250 billion API calls made since early in its launch year.[1][21] Furthermore, Krutrim has ambitious plans to develop its own AI silicon chips, with a target for India's first AI silicon chip by 2026, designed to manage complex AI tasks and workloads.[1][21][22] This vertical integration, from chips to cloud to applications, is a key part of Krutrim's vision to make India technologically independent in the AI domain and to address the specific linguistic and cultural needs of the Indian population.[3][23][19][24] Ola itself plans to use the Krutrim AI Cloud, estimating annual savings of around INR 15 crore by migrating from other cloud providers.[25]

The launch of Kruti enters a competitive AI landscape, both globally and within India. While aiming to provide an alternative to global giants, Krutrim emphasizes its focus on Indian languages, cultural context, and affordability.[3][23][24][26] The company has also launched Krutrim AI Lab, described as India's first AI frontier research lab, to foster open-source AI innovation and collaborate with academia, startups, and developers.[27][28] This initiative includes open-sourcing some of its models to enhance AI adoption in India.[27] Despite its ambitions, the company acknowledges the challenges, including data scarcity and the need for efficient training methodologies suited to the Indian context.[27][24] However, by focusing on deep localization and a full-stack approach, Krutrim aims to carve out a significant space in the rapidly evolving AI industry, with Kruti representing its first major consumer-facing agentic AI product.[3][23][6]