India is placing a strategic emphasis on developing voice-first, vernacular Large Language Models (LLMs) to establish a unique position in the global AI landscape. This initiative is spearheaded by the India AI mission, recognizing the potential of these models to address the specific needs and linguistic diversity of the Indian population.

Abhishek Singh, the CEO of India AI Mission, has articulated this vision, emphasizing that focusing on voice-based LLMs in Indian languages will provide a distinctive advantage over global tech giants. This approach acknowledges the significance of voice as a primary mode of interaction for a large segment of the Indian population, particularly in rural areas, and aims to bridge the digital divide by making AI more accessible and user-friendly.

The IndiaAI mission has garnered significant interest, with nearly 200 proposals for building foundational AI models trained on Indian datasets, languages, and contexts. The government has allocated ₹1,500 crores to support this development, demonstrating its commitment to fostering indigenous AI capabilities. This initiative involves collaboration with Indian startups to develop and deploy functional models trained on India-specific data. Four Indian startups—SarvamAI, Soket AI Labs, Gnani.ai and Gan.ai—have been empanelled by the central government for the project focussing on the mission's foundational model pillar.

To facilitate AI development, the Ministry of Electronics and Information Technology (MeitY) has launched initiatives such as the AI Kosha, a non-personal datasets platform, and the IndiaAI Compute Portal, which provides access to graphics processing units (GPUs) procured under the IndiaAI Mission. The national compute capacity for AI-based applications and LLM development has already crossed 34,000 GPUs, with plans to add another 10,000 to 15,000 GPUs by the end of July 2025.

Sarvam AI, a Bengaluru-based GenAI startup, is set to open source its sovereign LLM under the IndiaAI Mission. This multilingual, voice-focused model is designed specifically for India's needs and is being developed entirely within the country, utilizing local talent and infrastructure. Sarvam AI has received the highest subsidy under the IndiaAI Mission so far—₹98.68 crore—towards accessing 4,096 Nvidia H100 GPUs over a six-month period. The company's models will be released under permissible open-source licenses, allowing researchers, developers, and entrepreneurs to build upon them.

While promoting innovation, the IndiaAI Mission also recognizes the importance of regulations to ensure that AI applications comply with India's legal framework, including the DPDP Act. However, the mission aims to avoid strict regulations like those in the EU, focusing instead on promoting innovation while preventing potential harm.

The development of voice-first, vernacular LLMs aligns with the broader vision of creating inclusive, ethical, and accessible AI that empowers local innovation and reduces reliance on foreign technology. By focusing on Indian languages and voice-based interactions, these models have the potential to unlock new opportunities in language tech, AI applications, and population-scale digital services, ultimately transforming India's digital ecosystem.