OpenAI 'GPT-5.5 Instant' reported default
- OpenAI said on May 5 that GPT-5.5 Instant became ChatGPT’s default model for all users, contradicting claims the switch surfaced only in third-party reporting. - OpenAI’s own product post said the model brings “smarter, more accurate answers,” while a system card said it was treated as “High capability.” - OpenAI’s help pages say paid ChatGPT users can still use the model picker, while Enterprise admins must enable GPT-5.5 Instant.
OpenAI did publish an official announcement for GPT-5.5 Instant, and it said on May 5 that the model became ChatGPT’s default for all users. That means the central claim in a May 22 ChatForest review — that the switch happened without an OpenAI release — does not hold up against OpenAI’s own product page and help documentation. OpenAI also published a system card for GPT-5.5 Instant on May 5 and updated ChatGPT help pages describing how the model appears across user tiers. ### Did OpenAI actually make GPT-5.5 Instant the default in ChatGPT? OpenAI said on May 5, 2026 that it was “updating ChatGPT’s default model” to GPT-5.5 Instant, available to everyone. The company’s product page described the model as delivering “smarter, more accurate answers” with clearer and more personalized responses. OpenAI’s research index also lists that release on May 5. OpenAI’s help center separately says “by default, all users have access to GPT-5.5 models.” That page adds that older chats now run on GPT-5.5 equivalents after earlier model retirements, and says paid tiers can manually choose GPT-5.5 Instant or GPT-5.5 Thinking in the model picker. (openai.com) ### What, then, is wrong with the ChatForest framing? ChatForest published its review on May 22 and said OpenAI had replaced GPT-5.3 Instant with GPT-5.5 Instant on May 5 with “no announcement, no warning.” But OpenAI’s own May 5 product page and system card show there was an official release tied to that date. (openai.com) ChatForest may still have correctly identified the date of the default-model change. (help.openai.com) The disputed part is the suggestion that the switch was undocumented. OpenAI’s materials show the company publicly described the rollout, even if some users did not notice it in the app at the time. That last point is an inference from the timing of the documents and the review, not a stated OpenAI position. (chatforest.com) ### What can be verified about performance claims like SWE-Bench and hallucination cuts? ChatForest said GPT-5.5 Instant scored 88.7% on SWE-Bench and reduced hallucinations by 52.5% versus prior models in its benchmark testing. Those figures appear in ChatForest’s review and homepage snippets surfaced in search results. OpenAI’s official GPT-5.5 Instant product page snippet available in search results does not include those exact numbers. (chatforest.com) Instead, it uses broader language about accuracy, clarity and personalization. Without opening a primary-source benchmark table from OpenAI that matches the ChatForest figures, those metrics should be treated as third-party claims rather than official OpenAI performance disclosures. ### What did OpenAI say about safety and access? OpenAI’s GPT-5.5 Instant system card says this is the first Instant model the company treats as “High capability” in its cybersecurity and biological-and-chemical preparedness categories, with safeguards applied accordingly. That is one of the clearest official statements attached to the release. OpenAI’s help pages also show access is not identical across products. (openai.com) The ChatGPT help page says GPT-5.5 is available to all ChatGPT tiers, while a separate Enterprise and Edu page says GPT-5.5 Instant is disabled by default for Enterprise workspaces unless admins turn it on. The Business models page says paid users can use Auto, Instant and Thinking options in the picker. (openai.com) ### So what should readers take away from this story? The dated facts are straightforward: OpenAI released GPT-5.5 on April 23, made GPT-5.5 Instant the default ChatGPT model on May 5, and published both a product post and a system card that day. ChatForest’s May 22 article did not uncover an unannounced switch so much as amplify a change OpenAI had already documented. The next place to check is OpenAI’s help center and model release notes. (help.openai.com) Those pages are where OpenAI is documenting model availability, legacy-model retirement windows and whether GPT-5.5 Instant is enabled by default in specific products. (openai.com)