India's Homegrown AI Models Face Rigorous Tests at AI Impact Summit: Rivaling Global Giants Like DeepSeek

India's AI Impact Summit in New Delhi showcased homegrown artificial intelligence models undergoing real-world evaluations.

Sarvam AI launched its groundbreaking 105-billion and 30-billion parameter models, optimized for 22 Indian languages and trained entirely in India.

Government Support Fuels Sovereign AI Drive

The IndiaAI Mission has been pivotal, offering subsidized GPU access to startups building indigenous foundation models.

Prime Minister Narendra Modi praised these innovations, highlighting India's shift from AI consumer to creator.

Historically, India's AI journey accelerated with the IndiaAI Mission in 2024, aiming to counter foreign model dominance.

Performance Claims and Challenges

Sarvam claims its models outperform DeepSeek in cultural relevance and efficiency, sparking independent tests by media like CNA.

Experts note the impact on sectors like healthcare and agriculture, where multilingual capabilities address local needs.

Other firms like Soket AI demonstrated Eka stack for coding in Indian languages, broadening the ecosystem.

Future plans include expanding compute infrastructure and sector-specific models under government white papers.

These developments position India as a leader in ethical, sovereign AI for the Global South.

Successful tests could reduce reliance on Western and Chinese tech, boosting economic sovereignty.