Preflight Checklist
What's Wrong?
Description
Auto-compaction silently fails with the following error when a conversation context grows large:
Error: Error during compaction: 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
Use Claude Code with the default model (no model configured in settings.json)
Have a long conversation until auto-compaction should trigger
Observe the error above
Expected behavior
Compaction should either succeed using the same standard-context model as the conversation, or gracefully fall back to standard context without requiring usage credits or user intervention.
Actual behavior
Compaction attempts to use a 1M context model regardless of the configured conversation model, which requires usage credits to be enabled. The session becomes unusable until the user runs /clear.
This is a new behavior as of a day or so ago.
Workaround
Running /clear restores the session. The error message suggests using --model to switch to standard context, but this requires hardcoding a model name, meaning the user won't automatically benefit from future default model upgrades. And since /model reports the default already, it's unclear if this will help.
Environment
Claude Code version: 2.1.156 (Claude Code)
OS: Windows 11
Model: default (claude-sonnet-4-6)
No model set in settings.json
Additional context
The compaction step appears to independently select a 1M context model regardless of the model used for the conversation. Usage quota on claude.ai is well below limits — this is a feature gate issue, not a quota issue.
What Should Happen?
context should auto-compact with no error message
Error Messages/Logs
Error: Error during compaction: 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
spend hours on a long claude code session until the ring context status graph is closed.
Claude Model
Sonnet (default)
Is this a regression?
Yes, this worked in a previous version
Last Working Version
unknown. I don't think it's a model version issue.
Claude Code Version
2.1.156 (Claude Code)
Platform
Anthropic API
Operating System
Windows
Terminal/Shell
VS Code integrated terminal
Additional Information
No response
Preflight Checklist
What's Wrong?
Description
Auto-compaction silently fails with the following error when a conversation context grows large:
Error: Error during compaction: 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
Use Claude Code with the default model (no model configured in settings.json)
Have a long conversation until auto-compaction should trigger
Observe the error above
Expected behavior
Compaction should either succeed using the same standard-context model as the conversation, or gracefully fall back to standard context without requiring usage credits or user intervention.
Actual behavior
Compaction attempts to use a 1M context model regardless of the configured conversation model, which requires usage credits to be enabled. The session becomes unusable until the user runs /clear.
This is a new behavior as of a day or so ago.
Workaround
Running /clear restores the session. The error message suggests using --model to switch to standard context, but this requires hardcoding a model name, meaning the user won't automatically benefit from future default model upgrades. And since /model reports the default already, it's unclear if this will help.
Environment
Claude Code version: 2.1.156 (Claude Code)
OS: Windows 11
Model: default (claude-sonnet-4-6)
No model set in settings.json
Additional context
The compaction step appears to independently select a 1M context model regardless of the model used for the conversation. Usage quota on claude.ai is well below limits — this is a feature gate issue, not a quota issue.
What Should Happen?
context should auto-compact with no error message
Error Messages/Logs
Steps to Reproduce
spend hours on a long claude code session until the ring context status graph is closed.
Claude Model
Sonnet (default)
Is this a regression?
Yes, this worked in a previous version
Last Working Version
unknown. I don't think it's a model version issue.
Claude Code Version
2.1.156 (Claude Code)
Platform
Anthropic API
Operating System
Windows
Terminal/Shell
VS Code integrated terminal
Additional Information
No response