[BUG] Sonnet-only usage consuming both "Sonnet only" and "All models" limits simultaneously

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?

Observation: Usage Limit Tracking Inconsistency

I've noticed something interesting in my usage dashboard that I wanted to bring to your attention—I suspect you already know about it, but wanted to flag it from a power user's perspective.

My current metrics show:

• All models: 22% consumed
• Sonnet only: 15% consumed
• Usage pattern: Exclusively Sonnet

This suggests Sonnet consumption is being counted against both buckets. I understand the November 24 announcement positioned Sonnet with its own independent limit, which is strategically smart—it gives users like me more flexibility and encourages deeper engagement with the platform.

What I'm flagging

The current tracking doesn't match that positioning. I've seen similar reports from other power users (#14362, #23690), which tells me this is probably a known technical detail you're working through.

Why I'm mentioning it

I'm a Max 20x subscriber who plans heavy usage across multiple projects. I'm genuinely trying to optimize my consumption pattern and allocate intelligently between models. The inconsistency makes it hard to plan effectively—and I'd rather ask directly than guess.

Simple ask

Could you clarify the intended behavior? If Sonnet is supposed to have an independent limit (as the Nov 24 messaging suggests), that's actually quite elegant—it'd let me be more aggressive with Opus for complex reasoning while preserving Sonnet for routine work. If the current behavior is intentional, I can adjust my expectations accordingly.

I'm invested in Claude's ecosystem, so I'm flagging this as someone who wants to use it well.

Details

• Plan: Claude Max (20x)
• Current session usage: 25% (resets in ~45 min)
• Weekly limits active
• Related issues: #14362, #23690, #12487

What Should Happen?

If I use exclusively Sonnet, only the "Sonnet only" usage limit should deplete. The "All models" limit should remain unchanged since I haven't used Opus or other models.

Expected: Both limits should show ~15% consumed
Current: "All models" shows 22% while "Sonnet only" shows 15%

Error Messages/Logs

No error messages. This is a usage tracking discrepancy visible in the Claude.ai Plan usage limits dashboard.

Steps to Reproduce

1. Subscribe to Claude Max (20x) plan
2. Use Claude Code exclusively with Sonnet model (do not use Opus or other models)
3. Check your usage limits dashboard at claude.ai
4. Compare the "All models" percentage against "Sonnet only" percentage
5. Observe: Both percentages differ despite exclusive Sonnet usage

Claude Model

Not sure / Multiple models

Is this a regression?

Yes, this worked in a previous version

Last Working Version

Yes - The November 24, 2025 announcement stated "Sonnet now has its own limit" (independent), but current behavior shows it consuming both buckets

Claude Code Version

latest

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Terminal.app (macOS)

Additional Information

Related issues: #14362, #23690, #12487 show identical behavior reported by other users.

Screenshot attached: Dashboard showing usage discrepancy

Observation: This appears to be a systematic tracking issue affecting multiple users, not a one-off case.