Different versions of GPT reported by Copilot in VS Code

[informatik01]

Hi.

I have a GitHub Copilot subscription and have both the GitHub Copilot "main" extension (ID: GitHub.copilot) and its companion extension GitHub Copilot Chat (ID: GitHub.copilot-chat) installed.

Today I've just discovered that when you ask the same question about GPT version that GitHub Copilot is based on, it reports different versions depending on where you ask it in VS Code:

• in the dedicated chat view on the left - it reports "GPT-4"
• inline in an editor window - it reports "GPT-3"

Here is a screenshot for illustration:

🙋‍♂️ Is it a bug or is it a known "issue" or is it meant to be so?

---

Both VS Code and Copilot extensions versions are the latest at the time of writing this.

• VS Code Version: 1.89.1
• OS Version: macOS Sonoma 14.5

[digitarald]

Yes, I can confirm Copilot is multi-model and we keep running experiments to pick the most balanced models for performance and quality.

But models are generally bad at telling which version they are unless we tell them. So there is a bug you hit that inline chat is (most often) 3.5, not 3. So far most people asked in panel chat, which gets a model hint.

[informatik01]

@digitarald OK, thanks for the info.