Sarvam To Build Trillion-Parameter AI Model, Launches

What happened

Sarvam unveiled plans to build a trillion-plus parameter AI model in India and launched Sarvam Inference, an India-hosted inference platform. 2 and Gemma 4.

Why it matters

The startup also announced a San Francisco office and appointed Devendra Singh Chaplot as an advisor AI unicorn Sarvam today unveiled its biggest product roadmap yet, announcing plans to build a trillion-plus parameter frontier AI model in India while launching a domestically hosted inference platform as it seeks to build an end-to-end AI stack spanning models, infrastructure and enterprise software.

The Bengaluru-based startup made the announcements at its first developer conference, Epoch 2026, alongside a slew of launches across agentic AI, speech, vision and enterprise productivity. “We are very happy to announce that we are building a trillion-plus parameter model right here in India. We are building them from scratch to be competitive in coding, cybersecurity, simulation, science and more,” Sarvam cofounder Pratyush Kumar said at the event.

The AI unicorn did not disclose a timeline for the model’s launch. Sarvam also announced the opening of an office in San Francisco and appointed Devendra Singh Chaplot, who was part of the founding teams at Mistral AI and Thinking Machines Lab, as an advisor. Among the biggest launches at the event was Sarvam Inference, an inference platform that serves frontier open-source AI models using infrastructure hosted within India.

What to watch

2 and Gemma 4. The startup said Sarvam Inference is aimed at helping developers and enterprises access frontier AI models while ensuring inference workloads remain within India, an increasingly important consideration for enterprises and government organisations with data residency requirements.

Sarvam cofounder Vivek Raghavan described the move as part of the startup’s broader push towards “token sovereignty” – serving a larger share of the AI tokens consumed in India through domestic infrastructure rather than overseas cloud providers. The startup also said it has developed an agentic optimisation system that automatically improves model performance and claimed it delivers up to a 15X improvement in inference speed for certain models. Sarvam added that its developer platform now has more than 1 Mn registered developers.