BharatGen to launch 17B-parameter multilingual AI model at AI Impact Summit

The government-backed BharatGen programme will showcase its 17-billion-parameter 'Param2' model at the India AI Impact Summit 2026, focusing on Indian languages and public-sector use cases

BharatGen is set to launch a 17-billion-parameter multilingual AI model, called Param2, at the India AI Impact Summit 2026, which is set to kick-off on February 16 in New Delhi. According to the announcement, the model will support 22 Indian languages and is being positioned as part of India’s broader push to develop “sovereign” AI systems trained on domestic data and run on local infrastructure.

 

The model is being developed under the BharatGen consortium, which operates out of the Technology Innovation Hub at IIT Bombay and is supported by the Department of Science and Technology. The programme is part of the wider IndiaAI Mission, which is funding compute infrastructure, data platforms and model development for public-sector and national-scale AI projects.

BharatGen Param2: What it is

Param2 is described as a 17-billion-parameter “mixture of experts” model built to work across 22 Indian languages. According to the release, it has been trained on Indian-language data and is intended to be used across text, speech and vision tasks, rather than only as a text-based language model.

 

BharatGen said the model will be demonstrated at the summit through sector-focused applications developed with government and industry partners. These include use cases in governance, healthcare, education, cultural digitisation and financial services. Examples cited include a government-facing system for urban development and revenue departments, healthcare applications for doctor–patient interaction, and tools for digitising and accessing archival and cultural documents.

 

BhartGen also said the underlying models are being developed to support reasoning, maths and code-related tasks, and are being trained using infrastructure made available under the IndiaAI Mission, along with data drawn from what it calls the Bharat Data Sagar repository.

Why multilingual AI models matter for India

The focus on a large, multilingual model reflects a broader policy direction that has been visible across multiple government-backed AI projects over the past few years. Public services in India operate across dozens of languages, and much of the information citizens interact with comes in a mix of formats, including text, scanned documents and speech.

Text-only, English-first systems are a poor fit for many of these workflows. Multilingual models are meant to reduce that gap by allowing the same system to work across languages, rather than relying on separate tools or manual translation layers. This is also why platforms such as BHASHINI and projects like Adi Vaani and BharatGen have focused on language coverage as a core design goal, rather than as an add-on.

 

In practice, this approach is aimed first at government and public-sector use cases, such as forms, advisories, helplines and document processing, where language and format diversity is the norm rather than the exception.

How this fits into India’s wider “sovereign AI” push

BharatGen’s Param2 model sits alongside a growing set of India-focused AI efforts that are being framed around domestic data, local infrastructure and language coverage. In earlier work, the programme has already released models for text, speech recognition, text-to-speech and document understanding, and Param2 extends this into a larger, general-purpose multilingual system.

In parallel, Indian startups have also been building models focused on local languages and use cases. Bengaluru-based Sarvam AI, for instance, has reported benchmark results for its speech and document models on Indian-language tasks, where it says its systems outperform several global models on specific Indic benchmarks. That work, like BharatGen’s, is focused less on general chatbots and more on tasks such as speech recognition, OCR and document understanding, where language coverage and script handling make a measurable difference.

Displaying BharatGen Logo.png. Page 1 of 3khalid.anzar@bsmail.i