Text-to-speech has become a fiercely competitive benchmark sport. Cartesia announced today that Sonic-3.6 — its newest voice model — ranks first on both Artificial Analysis TTS leaderboards, citing win rates of 92% in English, 91% in Spanish, and 86% in French against the next closest competitor model in blind head-to-head evaluations with native speakers.
The result follows what Cartesia describes as an architectural rethink started at the beginning of the year, working from first principles rather than iterating on the previous generation. The company has not published technical details of the underlying changes. Blind evaluation methodology used by Artificial Analysis scores preference across naturalness, prosody, and pronunciation rather than a single automated metric, which makes the head-to-head framing harder to dismiss as benchmark optimisation.
Sonic-3.6 is generally available today via the Cartesia API and can be trialled at play.cartesia.ai. The Artificial Analysis leaderboard updates continuously as new models enter the evaluation pool, so the ranking reflects performance at time of publication.
To stay across the latest in cloud, AI and enterprise tech analysis from Compare the Cloud, subscribe to our weekly newsletter at https://www.comparethecloud.net/newsletter
