Gradium AI launches enhanced tts model with high accuracy and speed

Gradium AI has launched a new default text-to-speech (TTS) model, achieving an 81.0% human-rated pass rate on a 500-sentence hard-case evaluation set across five languages. This model outperforms competitors like Cartesia Sonic 3.6 (75.1%) and ElevenLabs v3 Conversational (65.4%). The new model, available via Gradium’s API and Studio, delivers a 216 ms time-to-first audio at P50 on Coval, 170 ms faster than its predecessor. The company claims it has the lowest hard-case failure rate at sub-250 ms first audio with minimal variance. The evaluation set, open-sourced on Hugging Face, includes 100 items across 10 criteria in five languages, with strict human scoring. Gradium also reports a tight 30 ms interquartile range across 480 runs, compared to a 165 ms spread for Cartesia Sonic 3.6. While not the fastest model on Coval’s benchmark, Gradium’s combination of accuracy and speed positions it as a strong contender. Existing users require no action, as the model is automatically applied without disrupting current voice setups.