Sonic, developed by the Carteisa team, is a low-latency voice model aimed at providing realistic voice generation capabilities for various devices. This model utilizes an innovative state-space model architecture to achieve efficient and low-latency generation of high-resolution audio and video. Sonic boasts a mere 135ms latency, making it the fastest in its class. The Carteisa team is dedicated to optimizing AI efficiency, making it faster, cheaper, and more accessible. Sonic's release marks a significant step forward in real-time conversational AI and computing platforms with long-term memory, foreshadowing new experiences for AI in real-time gaming, customer support, and other fields.