The Indian government has identified 20 indigenous AI foundation model proposals for support under the IndiaAI Mission, including 12 large language models and 8 small language models, Parliament was informed on July 22, 2026. The shortlist includes Sarvam AI’s 30-billion and 105-billion parameter models, Gnani.ai’s Vachana text-to-speech system and BharatGen’s multilingual foundation model, Param2 17B MoE.
The disclosure to Parliament confirms that India’s sovereign AI push has moved from funding announcements to concrete model development, with named companies and parameter counts now attached to the IndiaAI Mission’s foundation model programme. Sarvam AI’s two large language models were trained entirely in India, while Gnani.ai’s Vachana can clone voices across 12 Indian languages using less than 10 seconds of reference audio. BharatGen’s Param2 17B MoE model is designed for governance, education, healthcare and agriculture applications, reflecting the mission’s focus on public-service use cases alongside commercial ones.
What Do the 20 IndiaAI Foundation Models Cover?
Of the 20 shortlisted proposals, 12 are large language models and 8 are small language models, spanning general-purpose reasoning, multilingual understanding and voice technology built specifically for Indian languages. This mix signals that the IndiaAI Mission is not chasing a single frontier model but instead backing a portfolio of specialised, Indic-language-first systems designed for sectors such as governance, education, healthcare and agriculture, where global foundation models trained primarily on English and Western data have historically underperformed.
What Does This Mean for Indian Businesses Using AI?
For Indian enterprises, the emergence of homegrown foundation models means cheaper, India-specific alternatives to relying solely on foreign AI providers for multilingual voice and text applications. A speech-to-speech model like Vachana, for instance, could let customer service and voice-assistant products support regional languages more affordably than licensing comparable global tools. Enterprises building AI products for Tier-2 and Tier-3 India, where English proficiency is lower, stand to benefit most from access to sovereign models trained on Indian language data.
Industry Reaction and Expert Commentary
The Parliament disclosure comes months after India hosted the AI Impact Summit 2026 in New Delhi, where officials and global AI leaders discussed India’s strategic AI ambitions. Policy commentators have noted that India cannot outspend global frontier AI investment from the likes of the US and China, and must instead focus on deepening backward linkages to frontier AI while strengthening forward linkages through its own applied models — a strategy the 20-model IndiaAI shortlist appears to reflect.
What Happens Next?
The identified proposals will now move through IndiaAI Mission’s funding and development support process, with models expected to reach pilot deployment across government and enterprise use cases over the coming months. Watch for further disclosures on funding allocation per model and timelines for public release or enterprise licensing of models such as Sarvam AI’s 105-billion-parameter system.
Frequently Asked Questions
How many AI models has the IndiaAI Mission identified?
The government has identified 20 indigenous AI foundation model proposals, comprising 12 large language models and 8 small language models, as informed to Parliament on July 22, 2026.
Which companies are behind the shortlisted models?
Named companies include Sarvam AI, with 30-billion and 105-billion parameter models; Gnani.ai, with its Vachana text-to-speech model; and BharatGen, with its Param2 17B MoE multilingual foundation model.
Why is India building its own AI foundation models?
India lacks frontier AI capability of its own and relies on foreign models to stay competitive; the IndiaAI Mission aims to build sovereign, Indic-language-first models for governance, education, healthcare and agriculture rather than compete directly on frontier model scale.
Leave a comment