Key Info

Command Code now supports setting the model and reasoning effort for each individual subagent, allowing the main session to stay on one model while subagents run in parallel on a different one.

Highlights

  • Per-subagent configuration of both model and reasoning effort
  • Main session and subagents can use different models simultaneously
  • Subagents can run in parallel on a separate model from the main session
  • Available to try now in Command Code