[Bug] Token usage spike with non-Opus models after Opus 4.7 release

[MeterH1]

Bug Description
Since the release of opus 4.7, my token usage has spiked, even if I'm not using opus at all, but models like sonnet. I went from barely reaching 6-10% during a regular session to reaching 6-10% on only a messages. On the Max20x, that is unexpected and, for the cost, not appropriate. Thank you.

Environment Info

• Platform: win32
• Terminal: windows-terminal
• Version: 2.1.81
• Feedback ID: 593d7a32-5910-4819-99e6-ff3fe12e064c

Errors

[]

[0xbrainkid]

This feels important because the regression is not "Opus costs more" but "the release of one model appears to have changed usage accounting for other models too."

If the reporter is seeing Sonnet sessions burn through budget at a post-Opus-4.7 rate, then there are a few plausible failure modes:

• the accounting path started using the wrong model weight table
• non-Opus traffic is being attributed to an Opus-era pricing/usage profile
• or a UI/session meter changed independently of actual billing and is now overstating burn

The clue is the shape of the change: from 6–10% over a normal session to 6–10% after only a few messages, without intentionally switching to Opus. That points to a cross-model accounting regression more than normal variance.

Would be useful to compare the displayed session burn with the raw provider-side token counts for the same turns, because right now users cannot tell whether this is a billing problem, a quota-meter problem, or both.