Does anyone know why the ‘thinking’ (white dot icon) time is variable on GPT5?

I'm using GPT to learn Thai (extract idioms, identify colloquial fixed phrases, that kind of stuff), and I've noticed that when pasting subtitles in a custom prompt the answers come back between 'instant' and 30 seconds.

Because I'm using it HEAVILY (think 100's of queries a day), it's very frustrating to wait 30 seconds before reading the answer. There's no explanation as to what's going on server side.

Does anyone know if there's a way to tell it to go into some kind of grok-equivalent 'fast mode' to prevent these waiting times?

Thank you!

Leave a Reply