Cartesia Launches Sonic-3.6: New TTS Model Dominates Speech Quality Rankings

Cartesia has released Sonic-3.6, advancing its real-time speech synthesis technology. The model achieves top rankings on both Artificial Analysis speech evaluation boards, with particularly strong performance on the Controlled Voice benchmark where comparability is standardized across models. Built on state space architecture rather than transformers, Sonic-3.6 delivers audio onset in under 90 milliseconds. The service operates as a cloud API with subscription pricing starting at $299 monthly, translating to approximately $49 per million characters. Comparatively, it undercuts competitors like ElevenLabs Eleven v3 at $100 per million characters while maintaining superior quality ratings. The model includes specialized controls for agent-based workflows and supports multiple languages including English and Hindi code-switching. Cartesia offers the technology exclusively through its hosted platform without open-source weights.