Hello, I am using gpt-4o-mini to run an inference on an input of about 23k tokens. The model’s responses are being cut off after about 100 tokens. Tried to use gpt-4o for the same task and it works as expected. max_tokens parameter is set to 16384 so that is not what’s causing the issue. This task has worked with 4o-mini before and is a new issue that started just 15 minutes ago. Tried using playground and openai python sdk and the same thing happens. Anyone facing the same issue?
Look at the API finish reason.
Maybe content_filter? That will be if it is reproducing trained content. Your “lyric search engine” AI won’t work.
You can also use the new parameter max_completion_tokens.