GPT-5.5 Instant cuts ChatGPT hallucination rates by more than half in medicine, law, and finance while stripping out the "gratuitous emojis" that made responses feel cluttered. The update went live May 5 for all users.
OpenAI's new default model replaces GPT-5.3 Instant, which arrived March 3. The company says the shift prioritizes accuracy and concision over personality quirks. In internal evaluations, GPT-5.5 Instant produced 52.5% fewer hallucinated claims than its predecessor on high-stakes prompts covering medicine, law, and finance. It reduced inaccurate claims by 37.3% on conversations users had flagged for factual errors. The model also posted stronger benchmark results. It scored 81.2 on the AIME 2025 math test, up from 65.4 for GPT-5.3 Instant. On the MMMU-Pro multimodal reasoning benchmark, it hit 76 versus 69.2.
OpenAI is simultaneously making ChatGPT more transparent about how it generates answers. The update adds memory source visibility across all models, showing users which past conversations, files, or Gmail data influenced a response.
Users can delete outdated sources or correct errors. The memory source feature rolls out to Plus and Pro subscribers on the web first, with mobile access coming soon. Free, Go, Business, and Enterprise users get it in the coming weeks.
Shared chats won't expose memory sources to recipients. For developers, GPT-5.5 is available through the API as "chat-latest." GPT-5.3 Instant remains accessible to paid users for three months before deprecation.
OpenAI knows model retirements can spark backlash. When GPT-4o was deprecated in February 2026, users who called the model their "best friend" signed petitions to keep it alive. The company is giving users a three-month transition window this time.
