# GPT-5.5 Instant becomes ChatGPT's default model > OpenAI replaces GPT-5.3 Instant with a more concise default and adds visible memory sources for personalized responses across consumer plans. Canonical URL: https://www.devobs.io/news/news-openai-gpt-5-5-instant-default/ By: Jonah Reed Published: 2026-09-06T11:58:54.623Z Updated: 2026-09-06T11:58:54.623Z Event date: 2026-05-05 Section: AI OpenAI made [GPT-5.5 Instant](https://openai.com/index/gpt-5-5-instant/) the default ChatGPT model on May 5. The release focused on shorter answers, stronger visual and STEM reasoning, improved factuality, and more selective use of web search. It also introduced memory sources, a product surface that shows some of the past chats, saved memories, files, or connected data used to personalize a response. ## Default-model changes reach every workflow OpenAI reported lower rates of hallucinated and inaccurate claims than GPT-5.3 Instant on internal evaluations, with the largest stated gains in higher-stakes topics. Those measurements are useful release evidence, but application owners should still replay their own conversations. A default change can alter verbosity, tool invocation, refusal behavior, formatting, and how often the model asks for clarification even when top-line accuracy improves. GPT-5.5 Instant also uses context from previous chats, files, and connected Gmail when those features are enabled and available. Memory sources let users inspect and correct some of that context. OpenAI cautions that the view may show the most relevant sources rather than every factor that shaped a response. ## Personalization needs a deletion test Teams supporting users around personalized answers should test what happens after a source chat is deleted, a saved memory is corrected, a connector is disconnected, or a temporary chat begins. The visible source list should not be treated as a complete audit log. Product documentation must distinguish data available to the model from data actually displayed as a source. For model migration, keep a fixed evaluation set spanning factual questions, image inputs, connected-data tasks, and deliberately stale memories. Compare final-task correctness and user control, not just prose quality. OpenAI retained GPT-5.3 Instant temporarily for paid users, providing a bounded rollback window while teams recalibrated. ## Source references -