Question about Configuring Copilot's Language Model

[Gwihwan-Go]

Hi there,

First of all, thank you for developing this fantastic Copilot SDK! It’s been incredibly useful—I’ve successfully integrated it via the command-line interface, enabling me to run various experiments programmatically.

I do have a couple of questions:

1. Model Configuration: Is there a way to configure or specify the underlying language model used by Copilot (e.g., model size, version, or other parameters)? If it is, how can I do it？

2. Default Model: What is the default model version used in the current Copilot SDK release?

Any guidance or documentation references would be greatly appreciated. Thanks in advance!

[AntoineGS]

See here for how to change the completion model.
The default is (or at least was very recently, updates are pushed daily without changelogs though) gpt-35-turbo, and you can change it to use gpt-4o-copilot.

[tpope]

• Model Configuration: Is there a way to configure or specify the underlying language model used by Copilot (e.g., model size, version, or other parameters)? If it is, how can I do it？

The only supported configuration option is github.copilot.selectedCompletionModel, which as of today doesn't do anything useful. See #2 (comment).

• Default Model: What is the default model version used in the current Copilot SDK release?

The default is dictated server-side, and it recently switched to gpt-4o-copilot. See this announcement.