Subagents do not show their model or intelligence level

Open 💬 2 comments Opened Jul 23, 2026 by JavierPiedra
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What version of the Codex App are you using (From “About Codex” dialog)?

Version 26.715.72028

What subscription do you have?

Pro

What platform is your computer?

Darwin 25.5.0 arm64 arm

What issue are you seeing?

When using subagents in Codex, I cannot see each subagent’s assigned model or intelligence level (for example, medium or xhigh), or its token consumption.
When I ask the orchestrator to assign a particular model or intelligence level, it says it cannot set those deterministically. This means I cannot know which configuration each subagent actually received or correlate unexpected token usage with it.
I have started reaching my token limit for the first time since Sol became available.

What steps can reproduce the bug?

Feedback ID: 019f911d-833f-7b82-b132-50c67a467909

What is the expected behavior?

Each subagent should show its actual model, intelligence level, token use, and whether the configuration was explicitly assigned or automatically selected. The parent task should show aggregated subagent token usage.

Additional information

Reproduction steps:
Start a Codex task.
Ask it to spawn one or more subagents.
Ask the orchestrator to set a particular model and intelligence level.
Observe it cannot guarantee that assignment.
Observe that the task UI does not show each subagent’s actual model, intelligence level, or token usage.
Session ID: 019f911d-833f-7b82-b132-50c67a467909

View original on GitHub ↗

2 Comments

github-actions[bot] contributor · 1 month ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #33772
  • #34370
  • #34665

Powered by Codex Action

runacodeitor · 23 days ago

Angel here, founder of Runa. When subagent model, reasoning level, and token use are hidden, an unexpected quota spike is impossible to attribute. At https://runacode.io, we’ve measured ~46% lower token cost on real runs. New users get $50 of Runa machine usage and bring their own model API key. Want to try it on a multi-agent task?