@@ -50,32 +50,31 @@ Override settings from `~/.codex/config.toml`:
5050
5151``` bash
5252# Set model
53- codex exec -c model=" o3 " " prompt"
53+ codex exec -c model=" gpt-5.6-sol " " prompt"
5454
5555# Set sandbox permissions
5656codex exec -c ' sandbox_permissions=["disk-full-read-access"]' " prompt"
5757
5858# Multiple overrides
59- codex exec -c model=" o4-mini " -c ' features.stream=true' " prompt"
59+ codex exec -c model=" gpt-5.6-sol " -c ' features.stream=true' " prompt"
6060```
6161
6262## Model Selection
6363
6464### Available Models
6565
66- Common models (subject to OpenAI availability):
67- - ` o3 ` - Most capable, higher cost
68- - ` o4-mini ` - Balanced performance/cost
69- - Default varies by Codex installation
66+ Model availability changes with OpenAI releases — do not hardcode model names in prompts or settings unless there is a specific reason.
67+
68+ - ** Default (recommended)** : omit ` model ` / ` -m ` entirely; Codex uses the ` model ` value from ` ~/.codex/config.toml `
69+ - Example current models: ` gpt-5.6-sol ` , ` gpt-5.5 ` (check ` ~/.codex/config.toml ` or the Codex release notes for what is currently available)
70+ - Retired model names (e.g., ` o3 ` , ` o4-mini ` , ` gpt-5.x-codex ` ) will fail or be silently migrated
7071
7172### Selection Criteria
7273
73- | Task Type | Recommended Model |
74- | -----------| -------------------|
75- | Complex planning | o3 |
76- | Quick review | o4-mini |
77- | Architectural decisions | o3 |
78- | Simple validation | o4-mini |
74+ | Task Type | Recommendation |
75+ | -----------| ----------------|
76+ | Planning / review / architectural decisions | Codex default (usually the most capable current model) |
77+ | Quick validation on a budget | A lighter current model, if one is available |
7978
8079## Configuration Hierarchy
8180
@@ -118,7 +117,7 @@ Provide a code review with verdict (PASS/CONDITIONAL/FAIL)."
118117
119118``` bash
120119codex exec \
121- -m o4-mini \
120+ -m gpt-5.6-sol \
122121 -s read-only \
123122 " Quick validation: Is this function safe? [code here]"
124123```
@@ -216,7 +215,7 @@ codex review [OPTIONS] [PROMPT]
216215| Option | Short | Description |
217216| --------| -------| -------------|
218217| ` --uncommitted ` | | Review uncommitted/staged changes |
219- | ` --config <key=value> ` | ` -c ` | Override config values (e.g., ` model="o4-mini " ` ) |
218+ | ` --config <key=value> ` | ` -c ` | Override config values (e.g., ` model="gpt-5.6-sol " ` ) |
220219
221220### Usage
222221
@@ -228,7 +227,7 @@ codex review --uncommitted
228227codex review --uncommitted " Focus on security vulnerabilities"
229228
230229# With model override
231- codex review --uncommitted -c ' model="o4-mini "'
230+ codex review --uncommitted -c ' model="gpt-5.6-sol "'
232231```
233232
234233### Key Differences from ` codex exec `
@@ -267,11 +266,11 @@ source scripts/codex-helpers.sh
267266# For general execution
268267PROMPT_FILE=$( codex_write_prompt " $PROMPT_CONTENT " " plan" )
269268OUTPUT_FILE=" $( codex_tmp_path ' codex-output.md' ) "
270- codex_run_exec " $PROMPT_FILE " " $OUTPUT_FILE " " read-only" " o4-mini "
269+ codex_run_exec " $PROMPT_FILE " " $OUTPUT_FILE " " read-only" # model arg omitted → Codex default
271270
272271# For code review (preferred for review phase)
273272REVIEW_OUTPUT=" $( codex_tmp_path ' codex-review-output.md' ) "
274- codex_run_review " $REVIEW_OUTPUT " " o4-mini "
273+ codex_run_review " $REVIEW_OUTPUT "
275274```
276275
277276### Helper Functions
0 commit comments