3 ms·I think they're claiming it's achieved by text models, not voice models, fwiw.by usef- 28d agoI think they're claiming it's achieved by text models, not voice models, fwiw.