> ## Content Index
> Fetch the complete content index at: https://globalfeed.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Cartesia's Sonic-3.6 tops both Artificial Analysis speech leaderboards
- URL: https://globalfeed.ai/en/cartesias-sonic-3-6-tops-both-artificial-analysis-speech-leaderboards/
- Published: 2026-08-28T02:39:12.000Z
- Updated: 2026-08-28T13:41:54.000Z
- Description: The company says it stopped fine-tuning the existing approach and rebuilt from the architecture up. The model leads both the Provider Voice and Controlled Voice arenas, but is in beta on Cartesia's API only.
- Author: GlobalFeed Editor
- Tags: Cartesia, speech synthesis, AI, benchmarks, x-cartesia, dil-en, elle

Voice technology company Cartesia has made Sonic-3.6, its text-to-speech model, generally available. The company says that in January it decided to stop fine-tuning the existing paradigm and rebuild from the architecture up, and that it passed its own best model within two months.

On independent evaluator Artificial Analysis's Speech Arena, the model ranks first on both boards. On Provider Voice it holds 1286 Elo, ahead of Speechify's Simba 3.2 (1238) and Alibaba's Qwen-Audio-3.0-TTS-Plus (1237). On Controlled Voice it holds 1144, above Cartesia's own Sonic 3.5 (1099) and ElevenLabs' Eleven v3 (1060). These rankings update live and the scores can shift.

The technical difference is architectural: Cartesia uses state space models rather than transformers. By the company's figure, time to first audio is under 90 milliseconds. The model supports more than 40 languages.

One detail on access: Sonic-3.6 is currently in beta and available only through Cartesia's own API. The company's documentation still lists Sonic 3.5 as the stable version, and partners carry 3.5\. The model is offered as a hosted service; the weights are not downloadable.

For details see [Cartesia's blog](https://cartesia.ai/blog?ref=globalfeed.ai) and the [Artificial Analysis page](https://artificialanalysis.ai/text-to-speech/models/sonic-3-6?ref=globalfeed.ai).