News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
Krutrim-1

Krutrim-1

Multilingual language model for Indian languages

Krutrim-1 is a 7 billion parameter language model built by Krutrim AI Labs. Released in January 2024, it focuses on Indian languages and English. The model aims to address data scarcity for native tongues through extensive training on Indic datasets.

Share on WhatsApp

  • Trained on 2 trillion tokens with focus on Indic languages
  • Released in January 2024 by Krutrim AI Labs
  • 7 billion parameter dense transformer model
  • Outperforms GPT-3.5 on several Indic benchmarks

What it is

Krutrim-1 is a foundation model designed to serve the linguistic needs of India. Developed by Krutrim AI Labs, this 7 billion parameter system was released in early 2024. It functions as a dense transformer decoder, capable of processing and generating text in multiple Indian languages alongside English. The project aims to reduce the gap in performance between major global models and those supporting regional dialects.

How it came about

Training for the model occurred between October and November 2023. The developers utilised a dataset containing information up to April 2023. Two versions were made public on 31 January 2024. The base model was trained from scratch, while the instruct variant underwent supervised fine-tuning on top of the base weights. This release marked an effort to provide equitable representation for native tongues.

Inside the model

The architecture is a standard auto-regressive transformer decoder. It contains 7 billion parameters, similar in scale to Llama-2. Training involved 2 trillion tokens, with a specific focus on the largest known Indic language dataset to mitigate data scarcity. The model supports instruction following through a separate tuned version. Tokenisation and alignment processes were adjusted to handle the diverse morphological structures of Indian languages.

What it can do

Evaluations show strong results on Indic benchmarks such as IndicCOPA and IndicQA. In these tests, it often outperforms GPT-3.5 and other specialised models like Airawata. On English tasks, it achieves parity with similarly sized competitors, scoring an average of 0.54 against Llama-2 Chat. Strengths include creative writing, summarisation, and translation across languages like Hindi, Tamil, and Bengali.

How to use it

The model weights are available under the Krutrim Community License. Users can download the base and instruct variants from Hugging Face or GitHub. Integration requires standard transformer libraries. The licence permits community use but may carry specific conditions regarding commercial deployment or redistribution, distinct from fully permissive open-source terms.

Safety and testing

The developers note that training data from the internet may contain biases or toxic language. Consequently, the model might amplify these issues or generate unsafe content if prompted. Human oversight is recommended for sensitive applications. While safeguards exist, the system can still produce socially undesirable text or inaccurate answers in certain contexts.

Questions people ask about Krutrim-1

What languages does Krutrim-1 support?

It is natively multilingual, with strong support for Indian languages including Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Tamil, and Telugu. It also performs competently in English.

Is Krutrim-1 free to use?

The model weights are available for download under the Krutrim Community License. This allows for community access and use, though users must adhere to the specific terms outlined in the licence agreement.

How large is the Krutrim-1 model?

It is a 7 billion parameter model. This size makes it comparable to other mid-sized open models like Llama-2 7B, allowing it to run on relatively accessible hardware while maintaining strong multilingual capabilities.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Krutrim-1 yet, so the facts on this page come from the sources listed above: huggingface.co/krutrim-ai-labs/Krutrim-1-instruct. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.