According to CoinMeta, Cartesia has released a new generation of real-time speech synthesis model, Sonic 3.6, which is now available for use. It supports 44 languages and has ranked first in various benchmarks on artificial and analysis. In a blind test using native voices, Sonic 3.6 outperformed Alibaba’s Qwen-Audio-3.0-TTS-Plus, Google’s Gemini 3.1, and Controlled. voice uses the same set of cloned voices, but Sonic 3.6 still came in first. This model can generate 136.1 characters per second, while ElevenLabs Eleven V3 generates 46.7 characters per second, and Gemini 3.1 generates 24.2 characters per second. The price is $49 per million characters, which is half lower than that of Eleven V3, but higher than $27.6 for Qwen and $10 for Simba 3.2.