Cartesia Raises $100 Mn, Launches Sonic-3 — A Breakthrough in Real-Time Voice AI

Cartesia AI, a voice AI startup, has raised a total of $100 million in funding from Kleiner Perkins, Index Ventures, Lightspeed Venture Partners, and NVIDIA, founder and CEO Karan Goel announced on social media.

The company previously raised $64 million in March 2025. Alongside the funding, Cartesia unveiled Sonic-3, its latest large-scale model for real-time, natural conversation.

According to Goel, Sonic-3 delivers “breakthrough naturalness,” capable of generating laughter, tone, and emotion with unprecedented realism. It supports 42 languages and boasts an end-to-end latency of 190 milliseconds—making it the fastest conversational model on the market.

What sets Sonic-3 apart, Goel explained, is its reliance on State Space Models (SSMs) instead of the widely used Transformer architecture.

“Transformers are like rewatching the entire conversation from the start before saying each new word,” Goel said. “SSMs are like humans, remembering the topic and vibe of the conversation—enough context to speak naturally without replaying everything.”

Goel added that he and co-founder Albert Gu, who pioneered the SSM paradigm through their research on projects like S4 and Mamba at the Stanford AI Lab, are now witnessing their work being adopted across the industry.

Cartesia was founded in 2023 out of Stanford’s AI Lab by researchers Karan Goel, Arjun Desai, Brandon Yang, and Albert Gu, alongside Chris Ré, a Stanford professor and MacArthur “Genius” grant recipient known for his influential work in artificial intelligence.

Cartesia’s technology already powers millions of monthly conversations for companies such as ServiceNow, Cresta, and Decagon.

Users can try Sonic-3 for free or book a demo through Cartesia’s website. Goel even offered a bold guarantee: “If we can’t make your voice AI better than what you’re using now, I’ll donate $5,000 to your chosen charity.”