Unable to change the model
I'm running the latest VSCode (1.98.2) with the prerelease Copilot extension.
These docs https://docs.github.com/en/copilot/using-github-copilot/ai-models/changing-the-ai-model-for-copilot-code-completion#changing-the-ai-model-for-code-completion are completely off, because I don't even have a "Settings" menu available after clicking the copilot icon (step 1).
There are different instructions here https://code.visualstudio.com/docs/copilot/ai-powered-suggestions#_change-the-ai-model but they also don't work - I don't have the "Change Completions Model" option, so I can't complete step 2.
What I do have is a "GitHub › Copilot: Selected Completion Model" setting, but it asks for a model ID, which I don't have, and don't know how to get.
I guess they simply switched everyone to 4o: https://github.blog/changelog/2025-03-27-gpt-4o-copilot-your-new-code-completion-model-is-now-generally-available/
Watching this as I have the same issue. If you found the answer, would you mind sharing it?
Thanks for the feedback here. If you update to the latest GitHub Copilot extension pre-release version you should see the GitHub Copilot: Change Completions Model available with only one model: GPT-4o. As mentioned in the blog post this model is now the default if you are running an up-to-date extension version.
Thank you @devm33. This does seem to be available after switching to the pre-release version of 1.294.1479 for the GitHub Copilot extension.
Thanks @devm33. How do I change to a non-default model in this version? Or was this feature removed?
@trymzet the models you have access to will be shown in that picker. For most users, that is only the GPT 4o model right now.
When are the other models coming back? I wanted to try the new Sonnet specifically.
@trymzet The menu under GitHub Copilot: Change Completions Model only applies to the model used for code completions in the editor. For chat use cases, many other models are available. See the docs here: https://docs.github.com/en/copilot/using-github-copilot/ai-models/changing-the-ai-model-for-copilot-chat
I know, but I'm interested in code completion. I guess you're trying to corporate-tell me that this feature has been removed and you guys are not currently planning to bring it back. Thanks for clarifying.
I guess the issue can be closed then.
🙂 Sorry didn't mean to be indirect. Yes the previous GPT 3.5T model for code completions is being retired and GPT 4o is the default now: https://github.blog/changelog/2025-03-27-gpt-4o-copilot-your-new-code-completion-model-is-now-generally-available/
Can we only use GPT4? Would prefer to use Claude but even though I enabled it in my settings, it does not seem to appear in the dropdown
@callmedukey it seems they removed that feature, now you can only use GPT4 for completions. See dev's comments above
Can we only use GPT4? Would prefer to use Claude but even though I enabled it in my settings, it does not seem to appear in the dropdown
Hi - what you have enabled is Claude for the Chat features - it has never been available for code completion.
@kasavin I don't think that's the case - the official documentations of GitHub and VS Code I linked in OP both explain how to change the model for code completion. So it was available before, unless both websites are misleading?
@kasavin I don't think that's the case - the official documentations of GitHub and VS Code I linked in OP both explain how to change the model for code completion. So it was available before, unless both websites are misleading?
Hi @trymzet - sorry if I am being unclear - I think there are two questions in this thread:
- Can you change the model for code completion?
- Which models are available for code completion?
We introduced the model selector for code completion when the GPT-4o Copilot model was in preview and before the 3.5 Turbo-based model was retired. That way, it was possible to switch between the two to use either one. Now that the 3.5 Turbo-based model is retired, only the 4o Copilot model is available for code completion.
The picker UI should be visible again - I just tested in VS Code stable:
But this particular picker for completion never had Claude as an option.
I also came here because documentation is confusing and the selector only has one option. I also want to change the code completion to a claude model, because in my experience it has better results for my project (flutter + dart). I understand you are concerned with latency, so maybe those models aren't fast enough for code completion
I completely understand this is different from ask/edit/agent models which have their own lists of models
It also appears that the model picker in chat / completion is now gone...?
Is Micro$oft trying to lock us into their preferred model instead of letting us decide what model to use?
Thank you all for the feedback. The list of available models might vary based on your Copilot subscription and model availability--and might change (and has been changing) over time as we evolve the models.
We've identified these for next steps:
- We're adding clarifying information to the UI as suggested here (tracked internally, PR in progress)
- Update the docs for parity (https://github.com/microsoft/vscode-docs/pull/8335)
Once these PRs are merged, we will proceed to close this issue. If you have any further feedback, please don’t hesitate to file a new issue. Thank you again!
The list of available models might vary based on your Copilot subscription and model availability
But this isn't true? I'm in an enterprise org with paid subscriptions and all models enabled, yet we still can only choose one option for the code completions model.
I am confused as to why you even have documentation explaining how to switch the code completions model at all if we only have one option? I spent all morning trying to figure out why I can't switch the code completion model to Claude until I finally found this thread...
This whole situation is very strange.
I have been relying on Copilot code completion for months now and have always been under the distinct impression that the model I chose in the Copilot Chat window, at the bottom, was ALSO the model being used for everything else related to VS, including code completions.
And the reason I was under that impression was because everything implies that.
If you click the GitHub Copilot button top right of Visual Studio and then click Settings, there is then an option to tick Enable Copilot Completions.
The direct implication of that menu structure and where it is positioned is that whatever model you have selected in the Copilot Chat window (which is the ONLY place I was aware you could even select an AI model until now) is the model that will be used for code completions.
I didn't even realize there was a separate setting with a separate dropdown menu hiding under Options->GitHub->Copilot until now, and in fact that dropdown doesn't even have anything selected at the moment in my case.
I cannot believe that this entire time GPT-4o has been handling code completions and I've basically paid for a Github Copilot Pro license for nothing, because I almost never chat with the AI, my primary use for it is to sit and watch me and make code suggestions, and handle code completion, without me asking it anything.
I was also under the impression that any interaction I did explicitly have with the model I had selected in the Copilot Chat window was contributing to that model's domain knowledge and helping it to be more effective when it makes suggestions and handles code completion.
Clearly not, since a completely different model is doing the code completions, which means there is no shared context, unless I'm missing something.
The conclusion seems to be even under the "Configure Code Completions" setting, you can't change it away from "gpt-4o-copilot". Which is disappointing, the completions for 4o are really shitty.
It's ok that we can't do it as of now, but when will we be able to change to other models?
Thank you all for the feedback. The list of available models might vary based on your Copilot subscription and model availability--and might change (and has been changing) over time as we evolve the models.感谢大家的反馈。可用模型列表可能会根据您的 Copilot 订阅和模型可用性而有所不同,并且随着我们不断改进模型,列表可能会有所变化(并且一直在变化)。
We've identified these for next steps:我们已经确定了下一步要采取的措施:
- We're adding clarifying information to the UI as suggested here (tracked internally, PR in progress)我们正在按照此处的建议向 UI 添加澄清信息(内部跟踪,PR 正在进行中)
- Update the docs for parity (updated completions model selection instructions vscode-docs#8335)更新 parity 的文档( updated completions model selection instructions vscode-docs#8335 )
Once these PRs are merged, we will proceed to close this issue. If you have any further feedback, please don’t hesitate to file a new issue. Thank you again!这些 PR 合并后,我们将继续关闭此 Issue。如果您有任何其他反馈,请随时提交新 Issue。再次感谢您!
It seems that we still cannot independently change the code completion model and can only use the GPT-4.1 model. Can you tell us if this feature will be added? If so, when will it be on the agenda?
I have similar issue, missing model picker on vscode Linux
Unlike in Windows, copilot chat UI in Linux doesn't have a select menu to change the AI model, why?