[BUG] Abnormal token consumption on Claude for Mac Pro plan — persisting after June 2 outage resolution - Reference issue #41930
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?
Reference issue #41930
Subject: Abnormal token consumption on Pro plan — June 4, 2026
Plan: Claude Pro
App: Claude for Mac desktop app
Date issue started: June 4, 2026 (coinciding with June 2-3 outage)
Issue: Token consumption is 10-20x higher than normal. A single project orientation prompt consumed 21% of my session limit in approximately 20 seconds. Prior to June 2, I routinely worked for several hours on the same codebase in the same app without hitting limits. Nothing has changed on my end — same app, same project, same workflow.
Specific example: Opening a project with CLAUDE.md (121 lines) and reading 2 memory files consumed 21% of tokens in under 20 seconds.
Expected behaviour: Same workflow previously used less than 5% of tokens for project orientation.
Request: Please confirm whether my account was included in the emergency quota reset issued for the June 2 outage, and investigate why token consumption remains abnormal on June 4 after the incident was marked resolved.
What Should Happen?
Token consumption should be proportional to
the amount of content processed. Reading a
121-line CLAUDE.md file and 2 small memory
files (~200 lines total) should consume
less than 3-5% of a Pro plan session window.
Prior to June 2, 2026, the same workflow
on the same project consumed well under 5%
for project orientation. I routinely worked
for several hours per session on this codebase
without hitting limits.
Error Messages/Logs
No explicit error message. Symptom is silent
abnormal quota drain.
Observed consumption rates:
- Single project orientation prompt
(read CLAUDE.md + 2 memory files): 21% consumed
- Previous documentation generation session
(before noticing the issue): 90% consumed
within 15-20 minutes
- Prior to June 2 outage: same workflows
consumed <5% for orientation, hours of
work available per session
No error codes returned. Usage meter in
Claude for Mac app shows rapid consumption.
Steps to Reproduce
- Open Claude for Mac desktop app (Pro plan)
- Open existing project folder:
enterprise Next.js application, ~100+
source files, with CLAUDE.md (121 lines)
and /docs/memory/ documentation files
- Send single prompt:
"Read /path/to/project/CLAUDE.md"
- Observe: ~21% of session token window
consumed for reading one 121-line file
- Send follow-up prompt referencing
2-3 additional documentation files
- Observe: session at 40-50%+ after reading
~300 lines of documentation total
Expected: <5% consumed for steps 3-6
Actual: 40-50%+ consumed for steps 3-6
Claude Model
Sonnet (default)
Is this a regression?
Yes, this worked in a previous version
Last Working Version
Claude for Mac Version 1.10628.2 (deee0a)
Claude Code Version
Claude for Mac Version 1.10628.2 (deee0a)
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
This workflow ran without issue for several
months prior to June 2, 2026. The abnormal
consumption began coinciding with the June 2
outage and has persisted after Anthropic
marked the incident resolved on June 4, 2026.
Nothing changed on my end:
- Same app (Claude for Mac)
- Same project and codebase
- Same prompting workflow
- Same Pro plan subscription
- No dependency upgrades
- No configuration changes
Claude for Mac Version 1.10628.2 (deee0a)
— this is the only version I have access to
as the Mac app auto-updates. I cannot confirm
what version was running prior to June 2 as
the app updates silently.
Note: I am using Claude for Mac desktop app,
NOT the Claude Code CLI. The claude command
is not installed in my terminal
(claude: command not found). All development
work is done through the Mac app's Code tab.
Not applicable — using Claude for Mac desktop
app Version 1.10628.2 (deee0a), not Claude
Code CLI. Reporting here because the GitHub
issue tracker is the only channel where
Anthropic engineers have been confirmed to
respond to token consumption issues
(see related issue #41930).
Support ticket also submitted via email to
support@anthropic.com for account-specific
quota review.
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗