Spark usage unreported

Resolved 💬 6 comments Opened Feb 14, 2026 by iharc-jordan Closed Feb 21, 2026
💡 Likely answer: A maintainer (etraut-openai, contributor) responded on this thread — see the highlighted reply below.

What version of Codex CLI is running?

102 alpha 6

What subscription do you have?

Pro

Which model were you using?

gpt-5.3-codex

What platform is your computer?

windows wsl ubuntu

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

wsl

What issue are you seeing?

Despite using 5.3 - codex, I am seeing large usage being reported for the 5.3 codex spark model. This is displayed in the codex cloud usage portal.

I am thinking that maybe spawned sub-agents are using the spark model or there is a bug in model usage reporting.

If/when spark usage is fully depleted, I will give an update if I get any warnings or errors.

What steps can reproduce the bug?

Using gpt 5.3 codex (high) as normal causes depletion of spark model usage.

What is the expected behavior?

_No response_

Additional information

_No response_

View original on GitHub ↗

6 Comments

iharc-jordan · 5 months ago

Side note, you guys are on fire with the development of the cli and codex overall 🙌

etraut-openai contributor · 5 months ago

Do you have any features enabled in your config that are still under development? We do not recommend enabling these. Some under-development features may consume significant token usage.

iharc-jordan · 5 months ago
Do you have any features enabled in your config that are still under development? We do not recommend enabling these. Some under-development features may consume significant token usage.

Yes, I do. Another possibility is the memory feature if it uses the spark model. I believe memory extraction/consolidation etc is separately orchestrated. My primary usage is fine, it is just the spark usage that is being drawn down unexpectedly.

iharc-jordan · 5 months ago

I have confirmed that this is caused by sub-agent usage. As long as this reverts to a standard model automatically or if I can configure sub-agent profiles without having to manually build the app then it should be fine.

etraut-openai contributor · 5 months ago

Yes, the memory feature will currently drain usage very quickly. We strongly recommend against enabling features that are under development. When it's ready for use, we'll mark it as "experimental" and make it available in the /experimental menu.

iharc-jordan · 5 months ago

That's no fun 🤣 I'm happy to burn through usage if it means testing the latest and greatest!