India Unveils Three Sovereign AI Models at India AI Impact Summit 2026

India on 18 February 2026 unveiled three sovereign artificial intelligence models at the India AI Impact Summit in New Delhi, marking a major step toward building indigenous AI capabilities and strengthening digital sovereignty.

The models were developed by Sarvam AI, Gnani.ai, and an IIT Bombay-led consortium under the BharatGen initiative.

1. Sarvam AI’s ‘Vikram’ LLM

Sarvam AI showcased a 30-billion-parameter Large Language Model named Vikram, inspired by space pioneer Vikram Sarabhai.

Features:

• Trained and deployed within India
• Designed for Indian languages and contexts
• Focus on governance and enterprise use

2. Gnani.ai’s ‘Vachana TTS’

Gnani.ai launched Vachana TTS, a multilingual text-to-speech model.

Capabilities:

• Voice cloning using less than 10 seconds of audio
• Supports 12 Indian languages
• Preserves tone, pitch, and speaking style
• Optimised for low-bandwidth conditions

Use Cases:

• Government services
• Customer support systems
• Enterprise applications

All data and models are hosted within India, supporting secure deployment.

3. BharatGen Param2 17B MoE

The BharatGen initiative introduced Param2 17B MoE, a 17-billion-parameter multilingual foundational model.

Key Points:

• Mixture-of-Experts architecture
• Optimised for Indic languages
• Built for sectors like governance, education, healthcare, agriculture, and enterprise

What Are Sovereign AI Models?

Sovereign AI models are systems that are:

• Trained on domestic datasets
• Hosted within national infrastructure
• Governed under national laws

They aim to ensure data security, linguistic diversity, and technological independence.