[BUG] Claude Cowork blocked by "Usage credits required for 1M context" — usage credits enabled, /model workaround non-functional, triggers at 7% session usage during compaction

Open 💬 5 comments Opened Jun 5, 2026 by Tombombadil81

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?

Bug Description
Claude Cowork blocks all requests with "API Error: Usage credits required for 1M context" even though:

  • Usage credits are enabled at claude.ai/settings/usage
  • Session usage is only at 7% of the weekly limit
  • I never requested 1M context

The error appears to trigger during context compaction mid-session, not at the start. The workarounds suggested in the error message do not work — typing /model does nothing in the Cowork interface, and the error persists despite usage credits being active and available.

Steps to Reproduce

  1. Open Claude Cowork
  2. Begin a session with Sonnet 4.6 (default model)
  3. Work through a normal session (website/app building)
  4. At approximately 7% of weekly usage, error fires mid-session during compaction
  5. Error: "API Error: Usage credits required for 1M context · turn on usage credits at claude.ai/settings/usage, or use --model to switch to standard context"

Expected Behaviour
Cowork should continue working normally within my plan limits. Standard context (200K) should be used by default unless I explicitly request 1M context. The /model workaround suggested in the error should actually work.

Actual Behaviour
Session is completely blocked. Neither enabling usage credits nor attempting to switch models resolves the issue.

Workarounds Attempted

  • ✅ Enabled usage credits at claude.ai/settings/usage — error persists
  • ✅ Typed /model in Cowork — no effect
  • ✅ Usage credits confirmed available and active

Environment

  • Product: Claude Cowork (web)
  • Model: claude-sonnet-4-6 (default)
  • Plan: Pro with usage credits enabled
  • Session usage at time of error: ~7% of weekly limit

Additional Context
This appears to be the same root cause as #62063, #63896, #64398 and others — a recent update defaulting to 1M context regardless of user preference, with the error firing before model selection can process. The compaction step appears to independently select a 1M context model.

I have spent €127 in usage credits this month building a website and Android app in Cowork. I am concerned that some of those credits may have been incorrectly consumed due to this bug rather than genuine usage. I would appreciate clarification on whether a refund is appropriate.

What Should Happen?

Cowork should work normally within Pro plan limits without requiring 1M context. Standard 200K context should be used by default. The --model flag and /model command should successfully override the context window before any billing check occurs

Error Messages/Logs

API Error: Usage credits required for 1M context · turn on usage credits at claude.ai/settings/usage, or use --model to switch to standard context

Steps to Reproduce

Open Claude Cowork on a Pro plan with usage credits enabled
Start a new session using the default Sonnet 4.6 model
Work through a normal task (e.g. building a website or app)
Reach approximately 7% of weekly session usage
Cowork triggers context compaction automatically
Error fires: API Error: Usage credits required for 1M context · turn on usage credits at claude.ai/settings/usage, or use --model to switch to standard context
Attempt /model in Cowork — no effect
Confirm usage credits are enabled and available at claude.ai/settings/usage — error persists
Session is completely blocked with no working workaround

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

v1.11187.4

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_

View original on GitHub ↗

This issue has 5 comments on GitHub. Read the full discussion on GitHub ↗