Add support for Reasoning mode on GPT-5.6 models

[Slava-256]

What's happening?

Newly released GPT-5.6 models support standard and pro reasoning modes in the Responses API, see

https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode

and

https://developers.openai.com/api/reference/resources/beta/subresources/responses/methods/create#(resource)%20beta.responses%20%3E%20(method)%20create%20%3E%20(params)%200.non_streaming%20%3E%20(param)%20reasoning%20%3E%20(schema)

I am guessing we will no longer get standalone "Pro" models.

Where does this happen?

Big-AGI Pro (big-agi.com)

Impact on your workflow

None

Environment (if applicable)

No response

Additional context

No response

[enricoros]

Update: working on this and #1159 with priority - I was traveling yesterday; full steam on now.

[enricoros]

@Slava-256 Support for Pro reasoning mode added for all 3 models. For Sol, I've also added a "Pro" variant which automatically recreates the standalone Pro model - however: do you think it's good to create the model, or shall the users simply "duplicate" Sol and set the reasoning to Pro?

[Slava-256]

@enricoros Thank you for the quick update!

I imagine right now it feels more natural to have pre-made Pro model variant, since it feels like a continuity from older models, but if OpenAI continues with Pro mode being a parameter for any new model, then having it as a parameter will become new normal.

Given that 5.6 introduced a distinct new naming scheme (Sol, Terra, Luna), it already feels new and different, so I would say it makes sense to make the jump to always using reasoning mode parameter to select Pro mode now, and no longer have separate Pro model variant, unless the user makes a duplicate themselves.

[enricoros]

@Slava-256 Very good points - the biggest reason to keep the Pro would be just for usability for new users I guess, so they can Beam right away the Pro/normal.*

I've also made Temperature work more dynamically: if you select effort None, then those models shall allow you to specify the temperature right away (the slider acts better).

And finally I've made Reasoning Mode + Effort (incl. Max) available through OpenRouter as well.

*: maybe I'll keep Pro for 5.6 just to try this out, for ergonomics, but very likely I'm with you and we should make the jump to parametric only - but I'll need a tutorial for new users.