[BUG] Fast mode silently switches to Opus 4.6 without updating VS Code model picker — unexpected extra usage billing

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

When fast mode is toggled, Claude Code automatically switches from the user's configured model (Sonnet 4.6) to Opus 4.6 — as documented. However, the VS Code model picker UI does not reflect this switch. It continues to show the previously selected model (e.g. "Default (Sonnet 4.6)") with a checkmark, giving the user no indication that the actual model in use has changed.

Running /context reveals the true state: claude-opus-4-6, high effort — none of which is visible in the UI. All tokens are silently billed to extra usage at fast mode rates ($30/$150 per MTok), completely outside the plan quota with no ceiling.

What Should Happen?

When fast mode triggers an automatic model switch to Opus 4.6, the model picker should immediately update to reflect the actual model in use. At minimum, a visible session-level warning should notify the user that:

• The model has been switched to Opus 4.6
• All tokens are now billed to extra usage at fast mode rates
• Effort level and extended thinking settings may have changed

Additionally, disabling fast mode via /fast should either revert the model and effort to their pre-fast-mode state, or clearly prompt the user to do so manually.

Error Messages/Logs

No error messages. The issue is silent by nature — the UI gives no indication anything has changed.

Steps to Reproduce

1. Open Claude Code VS Code extension
2. Confirm model is set to Default / Sonnet 4.6 via the model picker (checkmark visible)
3. Toggle fast mode
4. Open the model picker again — it still shows Sonnet 4.6 as active
5. Run /context — actual model is claude-opus-4-6
6. Send several messages assuming Sonnet covered by regular quota
7. Check billing — all tokens are charged at Opus fast mode extra usage rates

Claude Model

Not sure / Multiple models

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.92

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

WSL (Windows Subsystem for Linux)

Additional Information

Billing impact in my case: ~5 short tasks resulted in €4.59 in unexpected extra usage charges. The model picker showed Sonnet (low effort) the entire time.

Why this is especially dangerous vs. other model-picker bugs:

• Fast mode bills 100% to extra usage — it bypasses the plan quota entirely with no natural ceiling
• There is no indication of the model switch and therefore the billing consequence in the VSCode extension

Related issues:

• #19157 — same root cause (picker shows Sonnet, actual model is Opus), triggered on first install rather than via fast mode
• #26950 — Opus 4.6 effort level cannot be inspected or reliably controlled; high effort default causes ~5x token usage with no warning
• #30726 — effort level silently downgraded when interacting with effort selection UI
• #35789 — model picker label and actual resolved model are inconsistent

This is distinct from #19157 because the trigger is fast mode's documented automatic model switch, making the billing consequences more severe (extra usage only, no ceiling) and the required fix broader (fast mode must sync model state to the UI on activation and revert it on deactivation).