OpenAI ends auto-switch from Instant to Thinking for Plus and Pro
Plus and Pro now default to fast Instant; hard prompts won't silently upgrade to reasoning unless you pick it. Health connections also flip to allow-by-default.
Copy markdown
Instant is now the default
As of Sept 14, ChatGPT Plus and Pro no longer auto-switch from Instant to the Thinking (reasoning) model, and the web app's 'Higher intelligence' toggle is gone with it. Unless you manually select a reasoning model, every prompt now runs on the fast Instant path.
The silent-quality trap
The old router quietly escalated hard prompts to reasoning; now a tough debugging, planning, or analysis question can land on Instant and just look answered. If you lean on ChatGPT for real work, pin a Thinking / reasoning model by hand — automatic escalation now survives only for safety routing, not for difficulty.
Elsewhere: Health connections go allow-by-default
The same day, new ChatGPT Health plugin connections started inheriting your global plugin permissions instead of asking per action — OpenAI says 70%+ of Health users had already opted into that. If you connect data sources, check your connector defaults so nothing runs without a prompt.