When working on complex tasks, subagents are automatically triggered without any user consent, and the system forces the use of the Composer 1.5 model (at 2x request cost).
This behavior is unacceptable for several reasons:
-
I already have access to a 1x cost model (Opus 4.6), yet I am forced into a more expensive model without justification.
-
The Composer model does not consistently deliver better results; in some cases, it performs worse.
-
There is no option to disable subagents or control which model is used, removing any ability to manage cost or performance.
This lack of transparency and user control directly leads to unnecessary cost increases and a degraded user experience.
At a minimum, the following controls should be provided:
-
A clear option to disable automatic subagents
-
Full control over model selection for all operations
Forcing users into higher-cost configurations without consent is not acceptable. This issue needs to be addressed.