PSA Recommends SLMs Alongside LLMs for India’s AI Ecosystem

India should prioritise the development of Small Language Models (SLMs) alongside indigenous Large Language Models (LLMs) to ensure wider linguistic representation and cost-effective artificial intelligence deployment, according to recommendations from the Office of the Principal Scientific Advisor to the Government of India.

In a recent white paper, the office highlighted that SLMs are more affordable to train and operate compared with large-scale models. These compact AI systems can be customised for specific sectors such as agriculture, healthcare, education, climate research and urban governance. Their efficiency also makes them suitable for public-sector applications where energy consumption and infrastructure costs are key concerns.

At the same time, building indigenous LLMs remains important to ensure AI systems reflect India’s linguistic diversity and social context. Heavy dependence on foreign AI models could lead to under-representation of Indian languages and potential biases in downstream applications.

The government has already approved proposals from several startups to develop domestic AI models. Companies such as Sarvam AI are working on a sovereign large language model designed for Indian governance and public services.

Research initiatives like BharatGen are also developing multilingual and multimodal AI systems to support India’s rapidly expanding digital ecosystem and strengthen technological self-reliance.

Source: Business Standard