Empty responses using o1-preview

[matthewgertner]

I selected o1-preview as the model and turned off streaming. For short prompts, it seems to be fine. But when I pasted in a large block of JSON, I just get empty responses with the cost of the message in the footer (e.g. "o1-preview-2024-09-12 using 1269 tokens ~= $0.042075").

Any idea what might be causing this?

[Ryan526]

In your profile for o1 preview change token settings to fit the max o1 can handle. Reasoning tokens are using up everything so it spits out a blank response if you don't do this.

Max Context: 128k
Max Output 32k

[matthewgertner]

I could only see one setting for max tokens. When I reverted it to the default of 128k, it started working as expected.

Thanks!