›_ Codex Issues
Codex Issues / #40127

CLI statusline keeps showing fast mode as off when it is actually on

Open 💬 1 comment Opened Aug 22, 2026 by rafaself

What version of Codex CLI is running?

codex-cli 0.149.0

What subscription do you have?

Plus

Which model were you using?

gpt-5.6-luna

What platform is your computer?

_No response_

What terminal emulator and version are you using (if applicable)?

_No response_

Codex doctor report

What issue are you seeing?

TUI statusline keeps showing fast mode as off when it is actually on

<img width="1914" height="395" alt="Image" src="https://github.com/user-attachments/assets/67ee9024-03cb-4594-b37a-4f3d7fa752e2" />

What steps can reproduce the bug?

Inside Codex TUI:

  • Execute the slash command /statusline;
  • Turn fast mode on to be shown in the statusline;
  • Execute /fast on

What is the expected behavior?

Statusline should show fast mode on/off according to the actual fast mode status.

Additional information

Fast mode feature should be active in the codex configurations.

View original on GitHub ↗

1 Comment

charle-z · 5 days ago

I traced this against the exact reported release tag (rust-v0.149.0, release commit 758ef40) and current main.

The status-line state path itself is internally consistent in that release: set_service_tier() updates the configured tier, recomputes the effective tier, and refreshes the model-dependent/status surfaces; FastMode renders Fast on exactly when the effective tier is ServiceTier::Fast (priority). The release also already contains status_line_fast_mode_renders_on_and_off, which exercises off -> set fast tier -> on.

The reproduction step has a separate issue: /fast is a dynamic ServiceTierCommand, and ServiceTierCommand::supports_inline_args() is explicitly false. Consequently /fast on is not dispatched as the fast-tier command. The supported command is the toggle /fast (no on argument).

So I cannot reproduce a stale status-line bug from the current report: /fast on does not establish the state the report assumes. If bare /fast leaves the status line at Fast off, that would be a different reproducible path worth tracing (model catalog / effective service-tier resolution); the current source and regression coverage do not support patching the status-line refresh path.