Repository navigation
gpt-5.5 capacity errors and recovered session returns with only ~60-70% context left #27149
Description
What version of Codex CLI is running?
codex-cli 0.137.0
What subscription do you have?
Not sure
Which model were you using?
gpt-5.5
What platform is your computer?
Darwin x86_64 (macOS)
What issue are you seeing?
I am repeatedly seeing:
Selected model is at capacity. Please try a different model.
After switching models to recover the session, it often comes back showing only around 60-70% context left instead of what appears to be a fresh/full context window.
What steps can reproduce the bug?
- Use Codex CLI on macOS
- Run an active coding session with tool use
- Hit the capacity error on
gpt-5.5 - Switch to another model such as
gpt-5.4 - Observe that the recovered session often shows only about 60-70% context left
What is the expected behavior?
- The session should continue without capacity interruptions, or
- if recovery is required, the recovered session should not come back with unexpectedly reduced available context
Additional information
- Switching to
gpt-5.4allowed the session to continue - I also noticed another similar report shortly before posting, so this does not look isolated
Activity
github-actions commentedon Jun 9, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
martinmclee commentedon Jun 16, 2026
I?m seeing this issue right now and wanted to add a direct data point.
I can reproduce this repeatedly during normal usage with that model. Fallback to another model is required to continue.
Could this be tied to backend regional model-capacity routing even for non-cloud/local sessions?
martinmclee commentedon Jun 16, 2026
I?m seeing this issue right now and wanted to add a direct data point.
I can reproduce this repeatedly during normal usage with that model. Fallback to another model is required to continue.
Could this be tied to backend regional model-capacity routing even for non-cloud/local sessions?