OpenAI releases GPT‑5.5 Instant as new default AI model on ChatGPT

OpenAI has officially launched GPT-5.5 Instant, the new default model for ChatGPT, succeeding the previous GPT-5.3 Instant. The company asserts that this latest version significantly mitigates hallucinations and delivers more precise responses in everyday interactions. With its focus on improving user experience, GPT-5.5 Instant is poised to transform the way millions engage with the AI chatbot daily. According to OpenAI, GPT-5.5 Instant boasts 52.5% fewer hallucinated claims compared to its predecessor. Additionally, inaccuracies have been reduced by 37.3%, thanks to extensive user feedback. The organization emphasizes that this model is notably smarter and more adept at handling everyday tasks, demonstrating enhanced capabilities in analyzing images, tackling STEM queries, and discerning when to utilize web searches for better answers. Performance metrics reveal impressive advancements, with GPT-5.5 Instant achieving a score of 81.2 on the AIME 2025 math benchmark, a substantial increase from the previous score of 65.4. Similarly, it scored 76 points on the MMMU-Pro multimodal reasoning benchmark, up from 69.2, showcasing its improved proficiency in mathematical calculations and logical reasoning. Another significant feature of GPT-5.5 Instant is its ability to leverage context from past interactions, files, and Gmail, allowing for more personalized responses. OpenAI is also introducing a new “memory sources” feature, which enables users to understand the context behind their conversations. Users will have the option to review and manage these memory sources, including the ability to delete or edit outdated information, providing a tailored experience. The rollout of GPT-5.5 Instant is underway for all ChatGPT users as the default model, while GPT-5.3 Instant will still be accessible to paid users for an additional three months before its retirement.