AI Voice Leaderboard Changes Again: Cartesia Surpasses Qwen and ElevenLabs, Supporting 44 Languages

According to Dynamic Insight monitoring, Cartesia has released the next-generation real-time text-to-speech model Sonic 3.6, which is currently in open Beta. The model supports 44 languages and has simultaneously secured the top spots on Artificial Analysis's two TTS leaderboards.

Provider Voice conducted a blind test using each model's native voice. Sonic 3.6 outperformed Alibaba's Qwen-Audio-3.0-TTS-Plus, SpeechifyAI's Simba 3.2, and Google's Gemini 3.1 Flash TTS. Controlled Voice had all models use the same set of cloned voices to focus more on model comparison, where Sonic 3.6 still ranked first.

Artificial Analysis measured Sonic 3.6's generation speed at 136.1 characters per second, ElevenLabs' Eleven v3 at 46.7, and Gemini 3.1 Flash TTS at 24.2.

The price is $49 per million characters, which is less than half of Eleven v3's price, but still higher than Qwen's $27.6 and Simba 3.2's $10.