Can I routing-force GPT-5.3-Codex-Spark to subagents? Draining my 5.6 Sol High weekly limits on 20X plan
Hey everyone,
I'm currently on the 20x plan and hitting a massive usage wall. I'm primarily using GPT-5.6 Sol High for my core tasks, but it burns through my weekly token allocation by mid-week.
Meanwhile, my GPT-5.3-Codex-Spark separate limits are sitting completely idle and going entirely to waste.
Since Spark runs on a separate quota and has that insane Cerebras speed, I want to offload the repetitive work to it. Is there any way to explicitly configure or force the system to spawn Spark subagents for lower-level tasks like code exploration, file structure analysis, or test runs? By default, my subagents just seem to inherit the parent 5.6 Sol model and eat my main quota.
Has anyone successfully bypassed this or figured out a custom configuration/workaround in the CLI or settings to route subagents to Spark?
Appreciate any tips!