Sarvam launches Indus AI chat app with offline capability

The Indus app, which entered beta testing on Friday, serves as the primary interface for Sarvam’s newly announced 105-billion-parameter large language model. According to TechCrunch AI, the launch follows the unveiling of this model and a smaller 30-billion-parameter version at the India AI Impact Summit in New Delhi earlier this week. The application supports both text and voice inputs, with the capability to deliver responses in audio format, positioning it as a multimodal tool for a diverse user base.

A key differentiator for the Indus app, as reported by The Independent, is its offline functionality. This feature allows the AI to operate without an active internet connection, a significant technical achievement that addresses connectivity challenges in various regions and could expand access to generative AI tools beyond urban centers with reliable internet infrastructure. This development places Sarvam in a distinct category compared to cloud-dependent services like ChatGPT and Claude.

The strategic importance of the Indian market for AI is underscored by usage statistics from the dominant players. TechCrunch AI reported that OpenAI CEO Sam Altman recently stated ChatGPT has surpassed 100 million weekly active users in India. Furthermore, Anthropic disclosed that the country accounts for 5.8% of total global usage of its Claude model, a figure that ranks second only to the United States. Sarvam’s entry represents a concerted effort to capture a segment of this burgeoning user base by focusing on local context and accessibility.

Beyond the consumer-facing chat application, Sarvam is pursuing a broader enterprise and hardware strategy. At the summit, the startup announced partnerships with several major corporations. According to TechCrunch AI, these include a collaboration with HMD to integrate AI capabilities into Nokia-branded feature phones and another with Bosch for developing AI-enabled automotive applications. These initiatives signal a plan to embed its AI models across a wide spectrum of devices and industries, from low-cost mobile hardware to advanced automotive systems.

The launch positions Sarvam to compete not only on the basis of language and local relevance but also through its unique architecture. The company’s focus on building its own foundational models, rather than fine-tuning existing ones from Western providers, suggests an ambition to create AI that is deeply optimized for the computational constraints and linguistic diversity of the Indian subcontinent. The offline capability is a direct result of this tailored approach to model development.

What comes next for Sarvam will be determined by the performance and adoption of its beta-stage application and the execution of its announced partnerships. The company has not publicly disclosed a timeline for a full public rollout or detailed pricing plans for its services. The success of its venture will hinge on its ability to demonstrate superior performance in Indian languages and use cases compared to established global giants, while simultaneously scaling its infrastructure to meet potential demand. The company’s progress will be a significant indicator of the potential for regional AI models to compete in a market increasingly defined by global scale.