Question: Model configuration and Computer Use at the same time

[mjakl]

The documentation mentiones the following feature as part of Stagehand

Claude is able to reason about the browser state, while Stagehand is able to take actions on the page with GPT-4o-mini or a computer use model. Stagehand is even smart enough to know when to use GPT-4o-mini and when to use a computer use model, i.e. on iframe detection.

The part Stagehand is even smart enough to know when to use GPT-4o-mini and when to use a computer use model is quite interesting as I have to deal with many cookie banners that are often in iframes and hard on Stagehand (even with the recent iframe improvements).

Though, I cannot find how to configure this anywhere. As I read this, it's not specific to the MCP server, but a feature of Stagehand. I would now assume that I can configure a standard model that acts on the DOM/snapshot and a computer use model that works visually.

Did I miss it in the docs, or am I misunderstanding what this means and how it works?

Thanks!