Model silently downgraded from Claude Opus 4.5 to Claude Sonnet 4 without user notification (billing concern)

[zhichao-poper]

Describe the bug

I explicitly selected Claude Opus 4.5 (which charges at 3x premium rate) before starting my session, but the actual model running was Claude Sonnet 4.

When I asked the AI "what model are you?", it reported being claude-sonnet-4-20250514, not Opus 4.5.

This is a serious billing/trust issue because:

1. Users are being charged Opus rates (3x) for Sonnet service
2. There is no notification that a model downgrade occurred
3. Users have no way to know they're not getting the service they paid for

Evidence

The AI self-reported as:

我是 Claude (claude-sonnet-4-20250514)

But I had selected "Claude Opus 4.5" via the /model command before starting the session.

Affected version

0.0.399

Steps to reproduce the behavior

1. Launch copilot
2. Run /model and select "Claude Opus 4.5"
3. Start working with the AI
4. Ask the AI "What model are you?"
5. Observe that it reports being Sonnet 4, not Opus 4.5

Expected behavior

1. If I select Opus 4.5, I should get Opus 4.5
2. If for any reason the system cannot provide Opus 4.5, it should:
   - Notify the user before starting the session
   - Ask for confirmation before using a different model
   - Adjust billing accordingly

Additional context

This appears to be a silent model substitution without user consent. Premium request billing should match the actual model used. Users need transparency about which
model is actually handling their requests.

Severity

This is a billing/trust concern that affects user confidence in the service. Users paying premium rates expect premium models.