[Bug] Claude Code incorrectly billing against extra usage instead of plan quota
Bug Description
Subject: Claude Code returning 'out of extra usage' error despite plan usage being under limit
Hi Anthropic Support,
I'm experiencing a critical bug with Claude Code where I receive an API 400 error despite having significant plan usage remaining.
Environment:
- Claude Code version: 2.1.29
- Plan: Claude Max (5x)
- Models tested: Opus 4.5, claude-sonnet-4-5-20250929
Error message:
API Error: 400 {"type":"error","error":{"type":"invalid_request_error","message":"You're out of extra usage. Add more at claude.ai/settings/usage and keep going."}}
The problem:
The error says I'm out of extra usage, but my actual plan usage shows:
- Current session: 2% used (resets 2am Asia/Seoul)
- Current week (all models): 0% used (resets Apr 15)
- Current week (Sonnet only): 0% used (resets Apr 15)
I am well within my Max plan limits. The error occurs regardless of which model I select. Meanwhile, claude.ai in the browser works perfectly fine with the same account.
Key observation:
Claude Code appears to be drawing from the "extra usage" bucket instead of the standard plan usage bucket, even when the plan limit has not been reached. This forces users to either enable extra usage billing or be locked out of Claude Code entirely — despite having an active Max plan with usage remaining.
This seems like either a billing system bug or an undocumented policy that Claude Code bypasses plan limits entirely and requires extra usage credits. If this is intentional, it is not clearly communicated and feels deeply unfair to paying Max subscribers.
Request:
Please investigate why Claude Code is triggering extra usage errors when plan usage is clearly under the limit, and clarify whether Claude Code is intentionally excluded from Max plan usage quotas.
Thank you.
10 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
Got the same issue, try reinstalling the claude code again. will fix the issue. Worked for me.
curl -fsSL https://claude.ai/install.sh | bashSame here. Issue has been fixed after re-installing.
Solved here reintalling too
here's the fix, obfuscated it so anthropic wont know how it is done. https://github.com/vinzabe/opencode-anthropic-max-fix/tree/master
placeholder
Placeholder comment was added in error during research, please disregard.
Closing for now — inactive for too long. Please open a new issue if this is still relevant.
Affected version: Claude Code v2.1.160 | Max 20X plan | June 2026
Adding data to support this bug report. I am experiencing the exact same issue — usage credits charged despite plan showing 4–6% consumption.
My session data across 4 sessions:
| Session | Model | Cache Read | Cache Write | Total |
|---|---|---|---|---|
| 1 | Sonnet 4.6 | 286.3M | 7.2M | 295.4M |
| 2 | Sonnet 4.6 | 139.1M | 6.9M | 147.1M |
| 3 | Opus 4.8 | 576.3M | 27.1M | 605.4M |
| 4 | Sonnet 4.6 | 106.1M | 10.3M | 116.8M |
| Total | | 1.107B | 51.5M | 1.165B |
Plan gauge at time of charge: Weekly all models 6%, Sonnet only 4%, 5-hour limit 0%.
Result: €43.08 charged to usage credits with no warning, no threshold alert, and no consent moment. Searching my full session transcripts confirms zero system-generated billing alerts were delivered — the only signals were cryptic "overload" / "retry_error" messages with no billing language.
Additional issue: 185 orphaned authorization tokens accumulating (37 pages) that cannot be deleted — each new session triggers a fresh cache write of full project context, compounding the token inflation.
Also experiencing "Usage limit reached" + "Upgrade" prompt on an active Max 20X plan (plan renewed 1 June 2026, less than 48 hours ago).
This is a confirmed product bug causing real financial harm. Anthropic support ticket open: Conversation ID 215474548931290.
/cc #46917
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.