ChatGPT defaults to GPT-5.5 Instant to reduce hallucinations

ChatGPT’s default model changed on May 5, and that matters more than it sounds. Most people never touch the model picker. They just open ChatGPT and type. So when OpenAI swaps the default engine underneath that box, it is effectively changing the product for everyone at once. This time the pitch is simple — GPT-5.5 Instant should make fewer things up without getting slower. (openai.com) ###(openai.com) as the new default model for ChatGPT, replacing GPT-5.3 Instant for all users. The company’s release notes frame it as an everyday model upgrade, not a niche “power user” option — better answers, less fluff, and lower hallucination rates in the version most people actually use. (openai.com) ### Why does “default” matter so much? Because de(openai.com)a menu helps enthusiasts. A default model changes the experience for hundreds of millions of ordinary chats — homework help, travel planning, coding questions, image uploads, quick factual lookups. OpenAI says Instant is the “daily driver,” which is why small gains in accuracy or clarity compound fast at scale. (openai.com) supposed to be better at? OpenAI says the gains show up in factual reliability, clearer and more concise writing, image understanding, STEM questions, and web-search behavior. That last one is easy to miss but important — a model that chooses better when to search has another way to avoid confidently wrong answers. The company also says responses should feel more tailored, with improved personalization controls. (op([openai.com)# Is this about speed or accuracy? Basically both. “Instant” is OpenAI’s low-latency line — the fast model meant for normal back-and-forth use. The usual tradeoff is that faster models can be sloppier. This update is OpenAI trying to bend that curve a bit — keep the quick feel, but cut hallucinations enough that the default assistant feels safer for routine use. That is the whole product bet here. (openai.com)et GPT-5.5 by default in ChatGPT, but paid tiers still get more control. OpenAI’s Help Center says Plus, Pro, and Business users can manually pick GPT-5.5 Instant or GPT-5.5 Thinking in the model picker, while automatic switching can also be configured. Free users get the default experience rather than the full control panel. (help.openai.com)? This sits inside a bigger cleanup of ChatGPT’s model roster. OpenAI retired GPT-5 Instant and Thinking in ChatGPT on February 13, 2026, along with several older models, while older chats may now run on GPT-5.3 or GPT-5.5 equivalents. So this is not just one shiny launch — it is part of a broader push to simplify the lineup and steer users toward newer defaults. (help. ([help.openai.com)hy should businesses care? Because model choice is turning into an operational setting, not just a research curiosity. Enterprise, Edu, and Business customers have to think about speed, limits, access, legacy compatibility, and when employees should use a fast model versus a deeper “Thinking” one. A more reliable default reduces the need to train every user to pick the right model manually. (help.opena([help.openai.com) ### So what’s the bottom line? OpenAI is making a very practical claim here: the best way to reduce hallucinations is not only to launch smarter flagship models, but to improve the fast one everyone already uses. If GPT-5.5 Instant really does cut wrong answers while staying snappy, the upgrade will matter less as a headline and more as a quiet change in how trustworthy ChatGPT feels day to day. (openai.com)