GPT5 Reasoning effort = none + Responses API weirdness

[lazka]

When enabling the Responses API and setting "Reasoning effort" to "none" I still get a reasoning response. When "increasing" it to "minimal" I get less reasoning.

So I wonder if the mapping is broken and "none" is not "none" here, but instead the "medium" default?

[ediblackk]

There is no way to disable reasoning, not even for mini/nano versions.
From what I've read and test myself, let reasoning high and verbosity low. Set temperature to 1. This improved the performances in my tasks.
None probably makes the model use medium (as it is by default).

Edit : I am new to LibreChat, may I ask how did you manage to select GPT-5 ? Should be available from model selection or is there any special setting ?

[lazka]

None probably makes the model use medium (as it is by default).

If that is the case I'd suggest renaming the label in the UI from "none" to "default", or remove it all together and select "medium" by default. That would be even more clear imo.

may I ask how did you manage to select GPT-5

I just added them to the OPENAI_MODELS env var, and using the latest development build. Afair it does auto discovery of models, but haven't tried that in a while.