Cartesia Sonic
Overview
Cartesia's Sonic line of streaming text-to-speech models built on state-space-model (SSM) architecture, aimed at ultra-low-latency real-time voice and cloning for voice-agent developers; the current generation is Sonic 3.5, following Sonic 3 in October 2025.
Field checks and source notes
When sources disagree on a field, every reading is listed here with its own citation. We do not take a midpoint, invent a missing figure, or decide which side is right — follow the citation and judge for your case.
These field-level notes are recorded verbatim in Chinese in the source data. Open the Chinese version
Plain-language benchmarks
Practical facts
| Context length | Not disclosed |
|---|---|
| Max output | Not disclosed |
| Parameters | Not disclosed |
| Input | Not disclosed |
| Output price | Not disclosed |
| CNY | Not disclosed |
| Free tier | Free plan: 20,000 credits per month (about 27 minutes of output). |
| Output speed | Not disclosed |
| Release date | |
| License | Not disclosed |
Show technical data
Technical benchmark table
| Benchmark | Score | Max / unit | Source type | Source | Test date | Conditions |
|---|---|---|---|---|---|---|
| TTS preference tts-preference | 1210 | elo_like | Human preference | AA Speech Arena 质量 Elo(人类偏好投票,81 个模型池);数值经第三方转引,待按 AA 原榜复核 |
| Version | Release date | Status | Openness | Parameters | What changed | Detail | |
|---|---|---|---|---|---|---|---|
| Sonic 3.5Showing version | Active | Closed source | Not disclosed | GA in May 2026 with more natural pacing and multilingual gains; Artificial Analysis Speech Arena Elo ~1,210 per third-party reporting. A later Sonic 3.6 is mentioned by trackers, unverified here. | Detail | ||
| Sonic 3(sonic-3-2025-10-27) | Superseded | Closed source | Not disclosed | Shipped 2025-10-27: 42 languages, emotion and laughter control, ~190 ms median latency per the vendor changelog. | Detail | ||
| Sonic(初代) | Superseded | Closed source | Not disclosed | The original Sonic release; exact launch date not verified in this pass and left blank. | Detail |
Last updated: · Sources on detail page