India Backs 20 Homegrown Sovereign AI Models
2 min readIndia is placing a large bet on building its own foundation models. The government has selected 20 indigenous sovereign AI models for support under the IndiaAI Mission, made up of 12 large language models and eight smaller ones. Electronics and IT Minister Ashwini Vaishnaw announced the shortlist in Parliament on July 23, 2026.
What sovereign AI means here
A sovereign model is trained, owned, and operated domestically rather than licensed from a foreign lab. For India that goal is practical as much as political. The country has 22 official languages and hundreds of dialects, and most Western models handle them poorly. Home grown systems can be tuned on local languages, data, and context from the start.
The plan
The 20 selected efforts span 12 large language models and eight small language models, the compact systems suited to phones and low cost hardware. Companies, startups, and research institutions are involved, with Reliance among the firms pledging to work alongside the IITs, IISc, and global technology partners on multilingual AI, data security, and skilled job creation.
The models sit inside a wider national push. The IndiaAI Mission is funding Centres of Excellence, roughly 500 data labs, and fellowships for 500 PhD scholars, 5,000 postgraduates, and 8,000 undergraduates, with AI labs reaching Tier 2 and Tier 3 cities.
Why it matters
Sovereign AI is becoming a defining theme across Asia, from South Korea’s national model push to Japan’s state backed infrastructure. India’s version is notable for its scale and its multilingual focus, aimed at a population of more than 1.4 billion. The open question is compute: training 20 models demands the chips and data centers that India is still racing to build.
Watch which of the 20 ship first, and whether any reach the quality of the global frontier. Success would give India a homegrown alternative in a market so far dominated by American and Chinese labs.