'Most real-world use cases don’t need trillion-parameter AI': BharatGen CEO Rishi Bal on Param 2 launch

On February 16, BharatGen introduced Param 2, a revolutionary artificial intelligence model featuring 17 billion parameters, specifically designed to cater to India's unique linguistic and sectoral requirements. This launch positions the model as a competitive domestic alternative to global AI systems from giants like OpenAI and Google, and it was officially revealed at the India AI Impact Summit 2026. With the ability to support 22 Indian languages, Param 2 aims to bridge the gap in AI solutions for local businesses and communities. BharatGen's CEO, Rishi Bal, highlighted the model's capabilities during an interview with Business Today, stating, 'Param 2 launches today, that’s our 17-billion-parameter mixture-of-experts model, supporting 22 Indian languages, along with functionalities for coding, mathematics, and reasoning.' He emphasized that this model is a pre-trained checkpoint that developers can adapt further to meet specific needs. BharatGen, backed by government support, believes that specialized, smaller models can create a more significant real-world impact in emerging markets compared to massive, one-size-fits-all AI systems. Bal noted, 'A majority of business, government, and consumer applications don’t require a trillion-parameter frontier model.' He likened such a model to 'a thousand PhDs all bottled into one,' asserting that organizations often need expertise tailored to their local contexts. Distinct from many global models that predominantly rely on English-language data, Param 2 was meticulously crafted for multilingual applications within India. BharatGen has adopted a phased release strategy, intending to introduce additional checkpoints and fine-tuned versions in the near future. This model will also serve as a foundational layer for future voice and vision systems. Bal indicated that Param 2 would initially focus heavily on text processing, with plans to roll out separate versions for voice recognition and document understanding. The development of Param 2 required significant computational resources, supported by the government. BharatGen received one of the largest allocations under the India AI Mission, utilizing thousands of graphics processing units (GPUs). Bal explained, 'The GPUs that we’ve used so far from the AI Mission have gone into the training of Param 2,' noting that the project has been leveraging around 4,000 GPUs for over a month. Looking ahead, the company aims to scale its operations, with larger models already under development. 'Expect to see multiple sizes of models, including smaller ones for edge inference and progressively larger variants throughout the year,' he added.