# an hour or so ago i was asked in this app if i wanted to turn on a more expensive turbo mode. now... ## Retrieval Notes - Session id: `019d080b-ecf6-71e1-9f02-b2329c1da89f` - Source: `Codex raw session` - Last activity: `2026-03-19T21:43:50.542Z` - Model: `gpt-5.4` - CWD: `/Users/nickyoung` ## My Notes <!-- Add your notes here. This section is preserved across syncs. --> ## Conversation ### User an hour or so ago i was asked in this app if i wanted to turn on a more expensive turbo mode. now. you guys have made it very difficult to find out how to turn it off. how do i do it? where is the setting, i have looked in what i take to be the usual places but it is late ### Assistant I’m going to check the recent daily notes for context, then inspect the app’s local config and any bundled help/docs so we can find the exact toggle instead of guessing. ### Assistant I’ve started the vault checks and I’m now looking for where this app stores its own preferences, especially anything labeled model, reasoning, or turbo. ### Assistant I found one concrete local setting already: your Codex config currently says `model_reasoning_effort = "medium"`, which suggests turbo is tied to reasoning level/model selection rather than a standalone “turbo” switch. I’m narrowing down where the app exposes that in the UI versus config. ### Assistant The app support folders are there, and the local CLI config is not in a “turbo on” state right now. I’m checking whether the desktop app has its own separate preference file, because that would explain why the toggle feels hidden. ### Assistant I’ve confirmed the desktop app keeps a tiny Electron-style `Preferences` file, but it doesn’t expose any obvious turbo flag in plain text. Next I’m checking whether this is actually controlled by the model picker inside the chat composer rather than a Settings pane. ### Assistant I’m checking the recent session logs now to see whether the app recorded the exact wording of that prompt. If it did, we can match it to the corresponding control instead of hunting blindly through menus. ### Assistant As of March 19, 2026, I do not see a separate global “turbo” switch in the Codex desktop app’s local preferences. On your machine, the durable default is already set to medium reasoning in [~/.codex/config.toml](/Users/nickyoung/.codex/config.toml), so it does not look like the expensive mode is currently stuck on. What to do in the app: 1. Look at the composer bar where you type your message. 2. Click the model / effort control there, not the main app Settings pane. 3. Choose a non-`pro` model and set reasoning to `medium` or `low`. 4. If you see anything like `high`, `xhigh`, or a `pro` model, switch away from it. If the UI is being annoying tonight, the fallback is just to verify these two lines in [config.toml](/Users/nickyoung/.codex/config.toml): ```toml model = "gpt-5.4" model_reasoning_effort = "medium" ``` That is already what you have now. If the app still behaves as if turbo is on, fully quit and reopen Codex so the desktop UI reloads the saved default. I also checked the app’s own desktop preference file at [~/Library/Application Support/Codex/Preferences](/Users/nickyoung/Library/Application%20Support/Codex/Preferences), and there is no visible turbo toggle there. *Se vuoi evitare la modalita piu costosa, usa un modello non pro con reasoning medium o low.*