Humanness Index™ · TTS model
by Cartesia
Sonic 3 landed in October 2025 as a major step for Cartesia's state space model architecture.
Standings as of Jun 13, 2026, 01:49 UTC
A real arena clip: a cloned source voice reading a customer support prompt at phone quality.
Sonic 3 landed in October 2025 as a major step for Cartesia's state space model architecture. It brought sub 100 ms latency, expanded coverage from 15 to 42 languages including nine Indian languages, and added expressive controls for emotion and even laughter. Instant voice cloning works from about 10 seconds of audio, and the model ships with SOC 2, HIPAA, and PCI compliance for regulated phone work.
Sources: cartesia.ai, vapi.ai
Emotion and laughter tags, 42 languages, and 10 second cloning. In our 50 trial streaming benchmark Sonic 3 returned first audio in a median of 166 ms including network time.
Sources: docs.cartesia.ai
Standings as of Jun 13, 2026, 01:49 UTC
| Rank | Provider | Model | Humanness | Latency |
|---|---|---|---|---|
| #17 | ElevenLabs | Flash v2 | 59 | 226 ms |
| #18 | Cartesia | Sonic 2 | 44 | 159 ms |
| #19 | Cartesia | Sonic 3 | 27 | 166 ms |
| #20 | Gradium | Gradium TTS | 24 | 332 ms |
| #21 | Cartesia | Sonic | 0 | 116 ms |
The benchmark is open source. Suggest a model, read the methodology, or ask us to put your voice in the arena.
