Does gpt 5.4 not support "stop" sequences with the chat.completions api?

I’m getting “400 Unsupported parameter: ‘stop’ is not supported with this model”. How is it possible that such a basic functionality is not being supported? What am I missing here?

Actually, there is a stop parameter:

Up to 4 sequences where the API will stop generating further tokens. The returned text will not contain the stop sequence.

What behavior are you seeing?

I’m aware, which is why I don’t understand why it doesn’t work for the 5.x models? It works perfectly fine for the old 4o model, but not for 5.4 (getting the “400 Unsupported parameter: ‘stop’ is not supported with this model”

It says:

Not supported with latest reasoning models o3 and o4-mini

Looks like they need to update the docs.

I say in web access mode for a number of reasons but there are some strange behaviors I’ve seen from 5.4 recently

5.4 tried nesting in my background process through Codex App and would relaunch every single time I closed the app… 3 days ago.

Today it’s been tak ing a launch app, + end command types of requests, and launching the app and idling a lot today…

Now it likes to respond to me after it makes an image to critique on the web portal… I kinda like that behavior, and agreed to it - but the ‘not stopping when it should’ seems to be a recent behavior with that model specifically?

I dunno… I’ve seen the models get rather unruly a number of times before ^.^

Best bet is to see if the Codex dudes, dudettes can help you out with that stuff. Maybe you need a debugging agent.