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
imgChitrarth

Chitrarth

Multilingual vision-language model for ten Indian languages and English.

Krutrim AI Labs released Chitrarth, a vision-language model that processes images and text in English plus ten Indian languages. It uses the Krutrim-1 language base with a SigLIP visual encoder to support tasks like image explanation across under-resourced linguistic regions.

Share on WhatsApp

  • Supports ten Indian languages plus English.
  • Uses Krutrim-1 LLM and SigLIP visual encoder.
  • Evaluated on new BharatBench suite.
  • Released under Krutrim Community License.
  • Outperforms IDEFICS 2 on several tasks.

What it is

Chitrarth is a general-purpose vision-language model developed by Krutrim AI Labs. The name combines the words for image and meaning. It integrates a large language model with a vision module to process both pictures and text. The system supports English and ten major Indian languages, including Hindi, Bengali, Tamil, and Telugu. Its primary goal is to serve users in these linguistic communities by enabling image understanding and description in their native tongues.

How it came about

The model was presented in a paper titled Bridging Vision and Language for a Billion People. Authors include Shaharukh Khan and colleagues from Krutrim AI Labs. The work was associated with the NeurIPS Multimodal Algorithmic Reasoning venue. The developers built the system using code references from Transformers and LLaVA-1.5 projects. They introduced it alongside a new evaluation suite called BharatBench to measure performance in under-resourced languages.

Inside the model

The architecture uses Krutrim-1 as the base large language model. A SigLIP component acts as the visual encoder, connected via a two-layer multi-layer perceptron. Training relied on multilingual image-text datasets. This design allows the model to align visual features with textual descriptions in multiple scripts. The structure follows standard decoder-only transformer patterns adapted for multimodal inputs, linking pixel data to language tokens for coherent generation.

What it can do

Chitrarth handles tasks such as visual question answering and image captioning. On the BharatBench suite, it achieved varying scores across languages. For instance, English scored 87.63 on POPE, while Assamese scored 55.59. It outperformed IDEFICS 2 and PALO 7B on several academic multimodal tasks. Performance remained competitive on TextVQA and Vizwiz benchmarks. The model provides a baseline for future research in Indic language vision tasks.

How to use it

Weights are available under the Krutrim Community License. Users can download the code from GitHub and run inference locally using Python. An online web interface is provided through the Krutrim AI Labs platform for direct interaction. The repository includes scripts for setting up the environment and processing image files with text queries.

Why it matters

The release establishes a reference point for multimodal AI in Indian languages. By introducing BharatBench, the creators provided a tool to evaluate models on ten specific languages. This helps address the lack of resources for these linguistic groups in artificial intelligence research. It encourages further development of systems that can understand local contexts and scripts.

Questions people ask about Chitrarth

Which languages does Chitrarth support?

It supports English and ten Indian languages: Hindi, Bengali, Telugu, Tamil, Marathi, Gujarati, Kannada, Malayalam, Odia, and Assamese.

What is the underlying architecture?

The model combines the Krutrim-1 large language model with a SigLIP visual encoder and a two-layer multi-layer perceptron for integration.

How can I use Chitrarth?

You can access it via a web interface on the Krutrim AI Labs platform or download the weights and code from GitHub under the community license.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Chitrarth yet, so the facts on this page come from the sources listed above: huggingface.co/krutrim-ai-labs/Chitrarth. 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.