Not Just LLMs, India Also Looking To Build SLMs: IT Minister

IT minister Ashwini Vaishnaw said that India is looking to build small language models (SLMs) for solving “distinct” problems, alongside a large language model (LLM).

“We are looking at our own digital models as well as our own smaller models, which can (solve) focused problems, distinct problems,” Moneycontrol quoted Vaishnaw as saying. 

Speaking virtually at the Nasscom Technology and Leadership Forum 2025, the IT minister also said that the government is focussed on creating a “large number of non-personal and anonymised datasets” for training indigeneously-built foundational artificial intelligence (AI) models.

“Can we create a very strong research base? For that, we have started centres of excellence (CoEs), three of them (have) already started, and one more is in the budget,” added Vaishnaw.

He also underlined that structured data sourced from state-backed digital public infrastructure (DPI) will help India “stand out” in the coming months and years in the AI race. Vaishnaw noted that the Centre is leveraging AI across the India Stack, including for applications such as Digi Yatra, Aadhaar, and UPI. 

“And the results have been phenomenal. The power of DPI actually can get multiplied 10X, 15X, 100x and we use AI in here… that’s going to be a huge win for us because we have that base already,” he added.

During a virtual fireside chat, Vaishnaw announced that Japan has given a patent to UPI’s “gateway system”.

According to a separate report by The Hindu businessline, the IT minister also noted that the country is aiming to transition from a services-led economy (referring to IT services companies) to a product powerhouse across sectors such as semiconductors, AI and consumer electronics.

He said that India’s AI ambitions extend beyond mere service provision or application development. “We could have limited ourselves to being a use case capital, being an application service provider capital of the world, but we want to become much more than that,” he added.

The minister said that the Centre is chasing a comprehensive AI strategy, which includes developing indigenous foundational models, creating anonymised non-personal datasets for training, establishing centres of excellence for AI research, and integrating AI education into universities.

While noting that construction is already underway at five semiconductor units in the country, the IT minister revealed plans to develop 25 semiconductor products at these upcoming facilities. 

On AI governance, Vaishnaw reiterated that the Centre will continue to focus on regulation that does not stifle innovation. “We should take care of the harm that may be caused to society and control them, but we should not let innovation be stifled the way many other countries have approached the problem,” he added. 

This comes close on the heels of reports that the Centre has received at least 67 proposals for building a homegrown AI foundational model. The government has also reportedly received proposals for building 20 LLMs from homegrown AI startups such as Sarvam AI, CoRover and Ola’s Krutrim.