Repository navigation
Global new-thread model setting is missing and project reset inherits an unexpected provider #6539
Description
Activity
T3 Code starts every new Codex thread at Low. My
~/.codex/config.tomlalready hasmodel_reasoning_effort = "high". Changing the composer to High does not provide a reliable global default. This happens with GPT-5.6-Sol on macOS.Rechecked this on T3 Code Nightly
0.0.37-nightly.20260830.1227on macOS with Codex CLI0.151.0.My Codex config selects
gpt-5.6-solwithmodel_reasoning_effort = "high", while Codex’s bundled model catalog advertises Low as Sol’s default reasoning level.Current T3 source appears to resolve a missing project or thread reasoning option from that catalog default, materialize Low in composer dispatch, and forward it as an explicit
turn/startoverride. That means Codex never gets to apply the configured High default.Setting each project’s New threads model to Sol / High works around it, but there is still no global T3 new-thread default. This seems to narrow the issue to T3 using the catalog default rather than the effective Codex configuration whenever no explicit project or thread effort is stored.
I am preparing a focused contribution for this.
Proposed scope: add a nullable, server-scoped default model selection for new threads, including model options such as reasoning effort. Resolution will be explicit draft selection, then project override, then configured server default, then the existing carried/last-selected behavior, then provider default. An unavailable configured selection will remain visible and block sending instead of silently substituting another model. The setting will be configured per environment; web/desktop and mobile thread creation will both consume it.
This will preserve explicit project overrides and rely on the existing automatic-project-default cleanup rather than adding another migration.
Closing as resolved by #9754 — Settings → Projects now has machine-wide (All projects) model defaults with per-project inherit/reset.
Problem
The new sidebar exposes a per-project default model and reasoning effort under project settings, but there is no equivalent global setting for new threads.
Resetting a project's model override does not inherit the provider's configured default. In my case:
gpt-5.6-solwithhighreasoning effortgpt-5.6-terra/xhighClaude Fable 5/highThe global Text generation model setting is unrelated. It only controls generated text such as thread titles and source-control content.
Expected
Add a global New threads → Model setting, matching the project-level model and reasoning controls. Project settings should clearly support Use global default, and resetting a project override should resolve to that setting.
The fallback should work consistently across project groups and environments. If the selected global provider is unavailable in an environment, the UI should show that explicitly instead of silently selecting another provider.
Reproduction