I am trying to understand and control how Codex app Automations interact with Fast mode / service_tier.
As a user, I want to be able to use Fast mode for my live/manual Codex chats, because faster responses are useful when I am actively working. But I also want unattended scheduled automations to be able to run at normal/standard speed by default, because they do not need immediate response speed and can otherwise consume credits unexpectedly.
From OpenAI Support’s AI assistant, the current answer appears to be:
- Codex Automations inherit client/profile service_tier defaults.
- If service_tier = “fast” is active, scheduled automations may run in Fast mode.
- There is no documented per-automation field to force standard/non-fast mode.
- service_tier is separate from model and reasoning effort.
- There is no documented local log/DB/billing field that clearly proves service_tier per automation run.
This creates a practical billing/control problem.
The desired behaviour is:
- Manual/live Codex chats can use Fast mode when I choose it.
- Most scheduled automations run standard/non-Fast by default.
- Specific automations can explicitly opt into Fast mode when appropriate.
- The UI/logs make it clear which service tier each automation run used.
Feature request:
1. Add an automation-level service_tier setting:
-
standard
-
fast
-
inherit profile/default
2. Show the service tier visibly in the Automations UI for each automation.
3. Add run-level visibility showing whether a given automation run used standard or Fast mode.
4. Document whether cron/project/thread heartbeat automations inherit global/profile Fast mode, and whether behaviour differs by automation type.
5. Add a global option: “Default automations to standard mode regardless of manual composer Fast toggle.”
Question:
Can someone please tell me how to do this if it is possible? If it is not currently possible, can OpenAI confirm whether per-automation service_tier control is planned?