Specifying the LLM in Chat Notebooks Doesn't "stick"

Greetings!

I recently got access to ChatGPT’s GPT-4 API and I’m able to use in Mathematica 13.3’s Chat Notebooks. However, when I select “GPT-4” under the “Advanced Settings” > “Model” in the inline chat cell config, it doesn’t save for subsequent chat input cells. Is there anyway to make “GPT-4” the default model for new chat cells in a Chat-enabled Notebook?

Selecting the model in every new chat cell is tedious.

Try this, for now:

CurrentValue[$FrontEnd, TaggingRules] = GeneralUtilities`ToAssociations@CurrentValue[$FrontEnd, TaggingRules];
CurrentValue[$FrontEnd, {TaggingRules, "ChatNotebookSettings", "Model"}] = "gpt-4"

See Avery’s comment for a more comprehensive answer. My code snippet will set “gpt-4” once and for all (at the $FrontEnd level, which is persistent across frontend notebook sessions)

When editing the selected model you can either: edit at the cell scope (what you’re OP is mentioning) or edit at the notebook scope

To edit it for the selected notebook, you want to edit the model selection through the toolbar.

There’s a pull request to make this selection a bit more intuitive globally (AKA a default state for any session) here: feature: Improve model menus by ConnorGray · Pull Request #304 · WolframResearch/Chatbook · GitHub
Before that gets merged and pushed out to users, there’s a programmatic way you could set this so that any new chat notebook uses gpt-4
CurrentValue[ $FrontEnd, { TaggingRules, "ChatNotebookSettings", "Model" } ] = "gpt-4”

Strange. I don’t have that toolbar button you noted. Is there something I need to do to get it show?

Nice! This worked. Thank you, Arnoud.