Sarvam Launches Multilingual AI Platform ‘Samvaad’ to Power Business Conversations

New Delhi: Homegrown AI startup Sarvam AI has unveiled Sarvam Samvaad, a conversational AI platform designed to help Indian enterprises rapidly build, test and deploy intelligent agents that operate in 11 Indian languages.

The platform supports multiple communication channels, including telephone, WhatsApp, websites and mobile apps, allowing businesses to offer seamless customer interactions in local languages.

The company claims that with its capabilities, enterprises can transition from pilot projects to production-grade AI agents within days. Several leading Indian firms are reportedly already leveraging the platform at scale.

Samvaad’s core strength lies in its language processing engine. It can accurately handle complex phrases, alphanumerics and proper nouns, making it effective in handling real-world, often ambiguous, customer queries.

Additionally, it enables real-time analysis of customer conversations, offering actionable insights that can enhance business intelligence and decision-making.

This new platform comes shortly after Sarvam’s release of Sarvam-Translate, an open-weight translation model built on Gemma3-4B-IT.

Sarvam-Translate supports 22 Indian languages, including Hindi, Tamil, Assamese and Manipuri. Unlike general multilingual models, Sarvam-Translate is optimised for long-form translation and can manage content in diverse formats such as LaTeX, HTML and OCR-processed files, while maintaining syntactic and structural accuracy.

The model is designed to interpret slang, idioms, emojis, and code syntax, making it particularly robust for modern content formats like social media, academic documents, and software codebases.

Sarvam’s recent developments are part of a larger push under the IndiaAI Mission, a government-backed initiative to create sovereign foundational models for India. Sarvam was the first startup selected under this mission and recently introduced Sarvam-M, a 24-billion parameter hybrid large language model based on Mistral Small. This model currently supports 10 Indian languages, including Bengali, Gujarati, Kannada, and Malayalam.

Following Sarvam’s selection, the IndiaAI Mission recently onboarded Soket AI, Gnani.ai and Gan.AI to join the effort to develop indigenous foundational models.

IndiaAI Mission under the IT Ministry also announced the selection of 10 high-potential Indian startups for the IndiaAI Startups Global (ISG) Initiative. The list includes names like Staqu, SatSure and CoRover.

Also Read –

• Indian Railways Joins Hands with BHASHINI to Roll Out Multilingual AI Services for Passengers