The gpt models like sol set with reasoning_mode=pro through api are totally far more insane than the default standard mode models through gpt subscriptions in all tasks. Vaguely remember it was promised that Codex users would have pro on Codex but it hasn’t. That mode is undoubtedly very important to use Codex for scientific research, and coding tasks also benefits a lot from the feature. Current’s pro subscription cannot set reasoning_mode=pro in codex config (the server silently rejects and still returns mode:standard) is so weird.
Pro reasoning mode is documented for the API: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode. We'll share the request for an explicit equivalent in subscription Codex. We don't have a timeline to share.