To change reasoning effort for a Codex CLI run, pass --config model_reasoning_effort=<value> in the invocation. For example, the OpenAI Cookbook demonstrates low: codex exec ... --config model_reasoning_effort=low. The available values depend on the selected model, so check its documentation before choosing one.
How to set reasoning effort for one CLI invocation
-
Check which model Codex CLI is using and which reasoning-effort values that model supports.
As an Amazon Associate I earn from qualifying purchases.
-
Add the override to the command:
--config model_reasoning_effort=<supported-value>. For example:codex exec ... --config model_reasoning_effort=low.Recommended Free Tools
Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy. -
Run the command. If the CLI rejects the value or the setting does not behave as expected, consult help or documentation for your installed CLI version and selected model.
#1 Best Overall
Sale
Cracking the Coding Interview: 189 Programming Questions and Solutions- Careercup, Easy To Read
- Condition : Good
- Compact for travelling
The OpenAI Cookbook documents this form in an example that pins @openai/[email protected]; it is a concrete example for that version context, not confirmation that every later release has identical syntax or precedence. See Build iterative repair loops with Codex.
Which reasoning-effort value should you use?
OpenAI says supported values are model-dependent and can include none, minimal, low, medium, high, xhigh, and max. That list does not mean every model accepts every value. For example, the reasoning guide says GPT-6 Astra does not support none, and GPT-6.1 Sol does not support none or minimal; it identifies medium as GPT-6.1 Sol’s default.
| Effort choice | When it may fit | Trade-off described by OpenAI |
|---|---|---|
low |
When speed and token economy matter more than deeper reasoning. | Lower effort generally favors faster responses and lower token use. |
medium |
A balanced starting point for many workloads, if supported by the model. | Balances speed and token use against reasoning depth; it is not a universal default. |
high |
Harder reasoning tasks where more latency and token use are acceptable. | Higher effort allows more complete reasoning, with greater latency and token use. |
xhigh or max |
Demanding cases where additional latency and token use are justified, if supported. | Higher effort allows more complete reasoning; outcomes vary by model and task. |
These are qualitative trade-offs, not measured guarantees for a particular workload. Confirm the selected model’s supported values and defaults in the OpenAI reasoning guide.
Free tools Windows power users keep installed
One-click scans. No signup required.
Do not confuse the CLI setting with API parameters
For Codex CLI, the documented configuration key in the Cookbook example is model_reasoning_effort. OpenAI’s API uses different names: Responses requests use reasoning.effort, while Chat Completions requests use reasoning_effort. These are settings for different interfaces; do not substitute API request syntax for the CLI option. See the reasoning guide, Codex Cookbook example, and Using GPT-6 migration guidance.
The migration guidance also describes a configuration_update input item for changing effort on later turns in supported standard, single-agent mode with successive GPT-6-family Responses API requests. That is an API mechanism, not the one-off Codex CLI override.
What if the override does not work?
-
Verify the exact value against the selected model’s documentation; a value listed for other models may not be accepted.
-
Check the installed CLI version’s help and documentation if the command rejects the option or the behavior differs from expectations.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy. -
Do not assume a universal precedence rule between this flag, saved configuration, environment variables, or aliases. The cited CLI example does not establish one for every release.
Quick Recap
SaleBestseller No. 2SaleBestseller No. 4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
