Hey, there’s no news yet. The behavior hasn’t changed: subagents can only use the model effort setting that’s currently saved in the model picker. We’re tracking this issue, but there’s no timeline yet.
For now, there are two workarounds:
- Before starting the chat, set the needed effort in the picker for each model family. The limit is at the family level, so you can, for example, keep GPT 5.6 Sonnet Medium and Grok XHigh at the same time. Fast counts too, so it’s best to turn it off if you don’t need it.
- Don’t specify a model for the subagent, and it will inherit the parent chat’s model.
You still can’t run two effort levels from the same family in one session, for example, Grok High and Grok Medium.
If you share where you’re working, IDE, CLI, or Cloud Agents, and which model combination you want to use, I can check whether it can be set up through the picker. I’ll post here when there’s an update.