Wrong model billed #3998

Describe the bug

When I prompted Copilot using auto model selection, it informed me that it used GPT-5.3-Codex in chat window:

But when done, it reported GPT-5.4:

Also when I opened my usage page in GitHub account settings, I saw GPT-5.4 as the only "Auto" model usage:

The thing is that the response style seemed to be that of GPT-5.3-Codex and I've been charged for GPT-5.4

Affected version

VSCode Insiders extension - GitHub Copilot Chat v0.56.2026070103

Steps to reproduce the behavior

1. Select "auto" model selection
2. Send prompt
3. Observe initial information on selected model
4. Observe response style
5. Observe summary information on model used
6. Observe AI usage billing chart in account settings

Expected behavior

Initial selected model info, summary model info, actually used model and billing info should all be unanimous

Additional context

No response