Cartesia text-to-speech models & pricing
5 models listed · appears on 3 of 3 boards · from $49.00/1M
Builds the Sonic family of low-latency text-to-speech models for real-time applications.
Models on the boards
ModelReleased$/1M charsBoardsComposite
Sonic 3.6Aug 2026$49.02/382.0
Sonic 3.5May 2026$49.03/376.0
Sonic 3Oct 2025$49.03/333.0
Sonic 2——1/34.0
Sonic——1/30.0
Latency from public benchmarks
Time to first audio, in milliseconds, as each benchmark publishes it. We don't measure latency ourselves, and the sources measure differently (harness, region, sample size), so each figure stays under its own source. Not part of the composite score.
ModelCoval, Time to first audio, p50, us-east-1, millisecondsSpeko, Time to first audio, p50, US East, millisecondsHumanness Index, Time to first audio, median, milliseconds
Sonic 3.6285 ms7-day window to Oct 11, 2026120 msmeasured Aug 28, 2026—
Sonic 3.5281 ms7-day window to Oct 11, 2026121 msmeasured Jul 3, 2026128 msmeasured June 2026
Sonic 3——166 msmeasured June 2026
Sonic 2——159 msmeasured June 2026
Sonic——116 msmeasured June 2026
Coval: Time to first audio, p50, us-east-1, read Oct 11, 2026 · Speko: Time to first audio, p50, US East, read Oct 11, 2026 · Humanness Index: Time to first audio, median, read Oct 11, 2026. How latency is sourced.