ChatGPT Plus and Pro Retire Automatic Switching to Thinking
Users must select an available reasoning level themselves; the web's Higher intelligence setting is removed and automatic safety switching remains.
Edited by Tyronne Panaino
OpenAI announced on September 14 that ChatGPT Plus and Pro are retiring automatic switching from Instant to Thinking globally. Users can still choose an available reasoning option themselves. The company is also removing the Higher intelligence setting from ChatGPT on the web for those plans.
The practical change is who selects the extra reasoning effort. A Plus or Pro user staying on Instant should no longer expect a complex request to trigger a higher thinking level automatically. OpenAI keeps an exception for automatic switching for safety purposes.
Choose the reasoning level before relying on it
OpenAI's current model guide says the Thinking slider controls the amount of reasoning on eligible paid plans. Available choices depend on the subscription and, for managed workspaces, workspace settings. Its table includes Medium and High for Plus; Pro also includes Extra High.
The guide says asking for more thought in ordinary language does not automatically change the selected thinking level on Plus or Pro. The actionable step is therefore to use the model controls when a task calls for deliberate reasoning, rather than assuming the wording of a request changes the setting.
For a recurring workflow, record the option you selected alongside the result. If you compare two runs, keep the source material and task requirements consistent, then inspect the answers against those requirements. This is an evaluation recommendation, not evidence that a particular level guarantees a better answer.
Separate the selection change from availability
Manual thinking remains subject to the account's allowance. OpenAI's guide says reaching a Thinking limit can cause ChatGPT to continue with another available model. Choosing a level and having remaining access are separate checks.
Before updating a team's instructions, verify the controls shown in the intended account. The September 14 announcement names Plus and Pro; it does not establish the same change for every managed workspace. A useful next checkpoint is consistent account-level behavior and any further clarification in the release notes.
Status
Confirmed product-policy announcement; medium internal confidence. Both fetched records are OpenAI documentation, with no independent testing of rollout behavior or answer quality.
Sources
Update note: Last reviewed September 15, 2026. Further documentation could clarify rollout behavior.
Sources
- OpenAI — ChatGPT release notes — official
- OpenAI — GPT-5.6 and GPT-6 Pro in ChatGPT — official
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.