Codex supports setting the reasoning effort via config overrides. You can set the model_reasoning_effort either in config.toml or as a one‑off CLI override with --config / -c. Here's an example:

codex --config model_reasoning_effort='"high"'