Set your default model and reasoning effort

configuration

// what it does

The model config key sets which model Codex uses; model_reasoning_effort tunes how hard it thinks, with medium the default and higher tiers (high, xhigh, max) trading speed and tokens for depth. Change either for the current session with /model, or persist them in config.toml.

// shell

$ /model
$ codex -c model="gpt-5.6-sol" -c model_reasoning_effort="high" "debug this race condition"

// gotcha

Higher reasoning effort is slower and burns more tokens, so reserve high and above for genuinely hard, multi-step problems. Check what is active with /status.

// resources