Cmd. /model
Switch the active model and, where supported, reasoning effort.
Key takeaways
/modelselects the active model and, in supported flows, reasoning effort, controlling the quality, latency, and cost profile of the thread.- The current default recommendation is
gpt-5.5, withgpt-5.4as a conservative fallback andgpt-5.4-minifor lightweight subagent work. gpt-5.2andgpt-5.3-codexare deprecated for ChatGPT sign-in and should not be new team defaults.- Since 0.124.0,
Alt+,cycles reasoning level without opening/model; Codex cloud tasks cannot change the default model, so keep Cloud policy separate. - Model availability depends on authentication, account, plan, and rollout, so use measured evals before a team-wide default change.
Original Command Declaration
/modelOfficial Summary
/model chooses the active model for the session and, in supported flows, exposes reasoning effort.
For team operations, treat it as the command that controls the quality, latency, and cost profile of
the current thread.
Usage
/modelGood Examples
- Use a stronger model before architecture, migration, or review work.
- Use a lighter model for bounded cleanup, search, or repetitive subagent tasks.
- Run
/statusafterward to confirm the active model and approval policy.
Similar Commands
| Command | Difference | Choose It When |
|---|---|---|
/model | Changes the model or effort | Task quality, speed, or cost needs a different model |
/fast | Changes service tier/speed | The model is fine but speed policy should change |
/status | Shows active model | You need to confirm before acting |
Use Cases
- Moving from exploration to deep implementation.
- Lowering cost for repetitive or low-risk turns.
- Standardizing a team model fallback when a preferred model is unavailable.
v0.123.0 to v0.139.0 Changes
GPT-5.5is the recommended default for most Codex tasks in the current OpenAI model guidance.- As of the 2026-06-12 re-check,
gpt-5.5is documented as available in Codex with ChatGPT sign-in or API-key authentication for CLI/SDK and app/IDE use. - The current recommended list centers on
gpt-5.5,gpt-5.4,gpt-5.4-mini, andgpt-5.3-codex-spark. gpt-5.2andgpt-5.3-codexare deprecated for ChatGPT sign-in and should not be used for new team defaults.- Codex cloud tasks currently do not let users change the default model, so keep Cloud task policy separate from
/modelpolicy. - Since 0.124.0,
Alt+,can cycle reasoning level without opening/model, which is useful for short-term depth adjustments. - Since 0.138.0, the TUI has fallback shortcuts for terminals that miss
Altbindings, and model-defined effort levels flow through in the order advertised by the model. - After 0.128.0, persistent
/goal, keymaps, remote-control, and longer-running workflows make model choice a question of work continuity and cost, not just single-turn quality.
Current Selection Notes (2026-06-12 KST)
- Default recommendation:
gpt-5.5 - Conservative fallback:
gpt-5.4 - Lightweight subagent or repeated work:
gpt-5.4-mini - Pro research preview or fast text-only iteration:
gpt-5.3-codex-spark - Remove from new defaults:
gpt-5.2andgpt-5.3-codexfor ChatGPT sign-in. - Do not standardize without evidence:
gpt-5.5-codex,gpt-5.5-mini, orgpt-5.5-nanoreferences are not part of the current official model guidance used for this handbook.
Cautions
- Model availability depends on authentication method, account, plan, and rollout.
- Use measured evals or task outcomes before making a team-wide default change.
Release Basis
- Judgment: confirmed addition
- First evidence version:
rust-v0.23.0 - Date (UTC): 2025-08-20
- Release link: https://github.com/openai/codex/releases/tag/rust-v0.23.0
Sources
- Codex slash commands: https://developers.openai.com/codex/cli/slash-commands
- Codex models: https://developers.openai.com/codex/models
- Codex CLI features: https://developers.openai.com/codex/cli/features