OpenAI began rolling out GPT-5.5 Instant on 5 May as ChatGPT's new default model. The company focused on everyday improvements: tighter answers, stronger factual performance, a more natural voice, and better use of relevant past context when personalisation was enabled.
Stable preferences that save repeated explanation.
Old context that changes the meaning of a new request.
Temporary Chat or an explicit reset of assumptions.
Why shorter answers were part of the release
Shorter is useful when it removes repetition rather than evidence. Personalisation can also save users from restating preferences, but it may bring old or unsuitable context into a new task. OpenAI added memory-source controls so users could see and manage some of the context involved.
What to test for yourself
Open the memory and personalisation settings before using the feature for important work. Remove outdated details, use Temporary Chat for a clean conversation, and state the current audience and constraints when an old preference should not apply.
See which memory features are enabled.
Correct details that no longer apply.
Compare accuracy and unnecessary length.
Verify current and high-impact claims.
The update aimed to make ordinary conversations feel less mechanical. That was a welcome change, but fluency and brevity were presentation qualities. Important facts still needed a source and a check.
