These models handle speech and help the interview avoid repeat questions. They run in your browser.
Avoiding repeat questions
Trained for this site · round 5
Checks whether your last answer already covers the next follow-up, so the interview can skip a repeat. It doesn’t grade your answers. Voice sessions on iPhone and iPad leave this model off to save memory.
Turning speech into text
base.en
Writes down your spoken answers. This is the English speech model used when you choose Speak.
Listening for pauses
v5
Detects when you start and stop speaking. It helps the app tell a pause from the end of your answer.
Reading questions aloud
nano · v0.8
Reads the interviewer’s questions when you switch on “Read questions aloud.” On iPhone and iPad, the site uses the device’s voice instead.
The final assessment is separate: when you choose Grade, your transcript is sent to server-side models to write the report.