API Stopped Working / "max_tokens" and "temperature" no longer allowed?

As of about 11am EST, all my scripts connected to openAI API stopped working - without change on my part. Errors showed that “max_tokens” and “temperature” no longer allowed. This both for o3-mini and gpt-4o. Is anyone else experiencing this? In addition, the API is painfully slow now - even with 4o.

Hi @drbo and welcome to the community!

I believe max_tokens was deprecated and instead max_completion_tokens parameter should be used. It was marked as deprecated for a while, but was perhaps supported anyway for gpt-4o until very recently, to give devs some time to switch over?

Regarding temperature - I believe this parameter is not supported by reasoning models, and it’s pre-set to some value (1 I think).

3 Likes