IIT Bombay’s BharatGen unveils 17B parameter sovereign multilingual AI model at India AI Impact Summit

Representative image

IIT Bombay-led consortium’s BharatGen unveiled its 17-billion-parameter multilingual foundational model, BharatGen Param2 17B MoE, at the India AI Impact Summit 2026 on February 18.

Backed by IndiaAI Mission and the Department of Science & Technology (DST), Param2 17B is positioned as a Mixture-of-Experts (MoE) model optimised for multiple Indic languages, with a focus on enabling AI adoption across governance, education, healthcare, agriculture and enterprise use cases.

How was Param2 17B trained?

The training pipeline for Param2 17B has been developed in collaboration with NVIDIA, leveraging NVIDIA AI Enterprise software and infrastructure. The stack includes NVIDIA NeMo open libraries such as NeMo-RL and Slurm workloads integrated with NVIDIA Base Command Manager to ensure scalability and performance.

BharatGen said the model has been built as part of its broader mission to develop India-first sovereign AI systems that reflect the country’s linguistic and cultural diversity. The initiative is aimed at reducing reliance on foreign foundational models while accelerating domestic AI capability.

Rishi Bal, CEO, BharatGen said, “We are extremely delighted at unveiling PARAM 2 17B at the  AI Impact Summit in Bharat Mandapam and also at the partnership with NVIDIA. As a truly sovereign AI initiative, the launch and unveiling of PARAM 2 17B marks a pivotal moment for India and we at BharatGen are elated at contributing to the country’s AI progress.”

Keeping the model open sourced

In a bid to encourage ecosystem participation, BharatGen will release the model, documentation and post-training workflows through its Hugging Face repository. This will enable developers, startups and enterprises to build, fine-tune and deploy India-centric AI applications.

The announcement comes amid India’s broader push to strengthen domestic AI infrastructure and capabilities, as global demand for generative AI systems continues to surge.

Who developed BharatGen?

Getting more than Rs 900 crores in funding, BharatGen is one of the biggest beneficiaries of the IndiaAI Mission 2025. This represents the single largest share sanctioned under the mission, underscoring BharatGen's central role in shaping India's artificial intelligence landscape.

IIT-Bombay led consortium includes IIT Madras, IIIT Hyderabad, IIT Kanpur, IIT Hyderabad, IIT Mandi, IIM Indore, IIIT Delhi and IIT Kharagpur. In November 2025, IIT Bombay registered BharatGen with the Registrar of Companies (ROC) as a non-profit organisation.

BharatGen has released a previous version of its LLM, Param-1, in July 2025 as a bilingual Large Language Model (LLM) with 2.9 billion parameters, pretrained on 5 trillion tokens in English and Hindi.

Building on this foundation, BharatGen went on to add multilingual and multimodal AI systems across all 22 scheduled Indian languages with the aim of ensuring equitable access to AI for every citizen, regardless of region or language.