Indian AI startup Sarvam has announced plans to build a 1-trillion-parameter artificial intelligence model, marking one of the most ambitious AI initiatives undertaken by an Indian company. Alongside the announcement, the company also unveiled significant upgrades to its Vision and Speech AI models, strengthening its focus on building advanced AI systems tailored for Indian languages and enterprise applications.
The announcements were made during Epoch 2026, Sarvam’s flagship AI conference held in Bengaluru.
India’s Biggest AI Ambition Yet
The proposed 1-trillion-parameter model will place Sarvam among a small group of AI companies globally attempting to build frontier-scale foundation models.
The company said the model is being designed to deliver stronger reasoning, multilingual capabilities, coding assistance and agentic AI features while supporting India’s growing AI ecosystem. It is expected to power applications across enterprises, government services, research and developers.
The announcement reflects Sarvam’s long-term vision of creating sovereign AI infrastructure built in India.
Vision AI Gets Major Upgrade
Sarvam also introduced improvements to its Vision AI platform.
The upgraded model offers better document understanding, handwriting recognition, table extraction and support for complex multilingual documents. It has been optimised to process scanned records, government documents, forms and business paperwork with higher accuracy.
The company says its Vision platform has already been used to digitise more than 35 million pages, making it one of India’s largest document intelligence platforms.
Speech AI Becomes More Capable
Sarvam’s Speech AI platform has also received significant enhancements.
The updated speech models deliver improved speech recognition, multilingual transcription and voice understanding across Indian languages. The company says its platform now processes over 500,000 hours of audio every month, supporting enterprise customer service, government programmes and voice-enabled AI applications.
The upgrades are aimed at making AI interactions faster, more natural and more accurate across India’s diverse linguistic landscape.
AI Platform Continues Rapid Growth
Sarvam highlighted the rapid adoption of its AI ecosystem.
According to the company, its platforms currently handle more than 2 million daily interactions and process approximately 10 million API requests every day. Its multilingual voice agents have been deployed in large-scale government initiatives, including programmes involving 17 million farmers, while enterprise deployments continue to expand across banking, insurance and public-sector organisations.
The company recently secured $234 million in Series B funding at a $1.5 billion valuation, providing additional resources to accelerate research and infrastructure development.
Building India’s AI Ecosystem
Sarvam’s latest announcements reinforce its strategy of building AI specifically for Indian requirements rather than adapting global models.
By combining frontier-scale language models with specialised Vision and Speech AI systems, the company aims to support businesses, developers and government agencies with AI tools capable of understanding India’s languages, documents and real-world workflows.
The planned trillion-parameter model represents a significant milestone in India’s efforts to develop globally competitive AI technology.
Summary
Sarvam has announced plans to develop a 1-trillion-parameter AI model while introducing major upgrades to its Vision and Speech AI platforms. The company says its Vision model has already digitised over 35 million pages, its speech platform processes more than 500,000 hours of audio every month, and its AI ecosystem now handles over 2 million daily interactions as it expands India’s sovereign AI capabilities.
