The Story

Indian AI startup Sarvam AI has announced two new large language models — Sarvam-30B and Sarvam-105B — aimed at improving AI capabilities across Indian languages and enterprise applications.

According to the official announcement from Sarvam AI, the models are designed to deliver strong multilingual performance while supporting applications tailored to India’s linguistic diversity.

About the new models

• Sarvam-30B: A mid-sized large language model optimized for efficient deployment across enterprise and developer workloads.
• Sarvam-105B: A larger frontier-scale model aimed at advanced reasoning, enterprise AI applications, and high-performance multilingual tasks.
• Indic language support: Both models are designed to handle multiple Indian languages with improved contextual understanding.

Focus on India’s AI ecosystem

Sarvam AI is positioning its models to support India’s growing demand for AI systems that can operate effectively across diverse regional languages and digital services.

The company says the models can support applications such as customer support automation, enterprise assistants, translation systems, and public sector digital services.

Why it matters

Most frontier AI models are primarily optimized for English and a small set of major global languages. Efforts such as Sarvam’s aim to close the gap for multilingual societies where hundreds of millions of users interact in regional languages.

The launch highlights how regional AI companies are building models tailored to local markets rather than relying solely on global foundation models.