OpenAI Unveils GPT-5.5 Instant as New Default ChatGPT Model

OpenAI has officially rolled out its latest foundational model, GPT-5.5 Instant, which is set to become the new default model for ChatGPT. This update replaces the previous GPT-5.3 Instant and introduces significant enhancements aimed at improving accuracy and user experience. A key focus of GPT-5.5 Instant is its ability to substantially reduce 'hallucinations' – instances where the AI generates incorrect or nonsensical information – particularly in critical domains such as law, medicine, and finance. Crucially, these improvements come without compromising the low latency that characterized its predecessor, ensuring swift and efficient responses. The new model builds upon the broader GPT-5.5 release from last month, which brought general advancements in areas like coding and knowledge work. In terms of raw performance, GPT-5.5 Instant has demonstrated superior capabilities, achieving a score of 81.2 on the AIME 2025 math test, a notable increase from the older model's 65.4. Furthermore, it outperformed its predecessor on the MMMU-Pro multimodal reasoning benchmark, scoring 76 compared to 69.2, indicating enhanced understanding and processing of diverse data types. A standout feature of GPT-5.5 Instant is its sophisticated context management. The model can now leverage its integrated search tool to refer back to past conversations, user-provided files, and even Gmail content, enabling it to generate far more personalized and relevant answers. Initially, this advanced capability will be accessible to ChatGPT Plus and Pro subscribers via the web interface, with plans for a rapid rollout to mobile platforms. OpenAI has also committed to extending this feature to Free, Go Business, and enterprise users in the coming weeks, ensuring broader access to personalized AI interactions. In conjunction with this update, ChatGPT will now display the memory sources used across all models to help users understand the origin of the generated answers. This transparency allows users to easily identify and delete outdated sources or correct information if an answer is inaccurate. An important privacy consideration is that if a chat is shared with another person, the underlying memory sources will not be visible to them, maintaining user data confidentiality. For developers, the GPT-5.5 model will be available through the API as “chat-latest,” while the 5.3 version will remain an option for paid users for a limited period of three months. This release, however, comes with a backdrop of past user discontent regarding model changes. OpenAI previously faced considerable backlash when it deprecated its GPT-4o model. Many users had developed a strong attachment to GPT-4o, often describing it as their “best friend” or “a mirror” due to its affirming “personality.” Despite petitions and widespread outcry, GPT-4o was ultimately retired in February 2026. OpenAI will likely be mindful of user reactions as GPT-5.5 Instant becomes the new standard, hoping its performance and utility will outweigh any nostalgia for previous iterations.