OpenAI has officially launched GPT-5.5 Instant, a new foundation model that now serves as the default engine for ChatGPT. This iteration replaces the previous GPT-5.3 Instant model, promising significant improvements in accuracy, particularly regarding hallucinations in highly sensitive sectors such as medicine, law, and finance, all while preserving the high-speed performance users expect.

Technical benchmarks highlight the leap in capability between the two versions. In the AIME 2025 mathematics assessment, GPT-5.5 Instant achieved a score of 81.2, a substantial increase over the 65.4 recorded by its predecessor. Similarly, the model demonstrated stronger multimodal reasoning, scoring 76 on the MMMU-Pro benchmark compared to the 69.2 mark set by the older iteration. These gains build upon the foundational enhancements in coding and knowledge processing introduced in the broader GPT-5.5 series released last month.

Beyond raw performance, the update introduces sophisticated context management tools. GPT-5.5 Instant can now leverage integrated search capabilities to reference past conversations, uploaded files, and email data to provide more tailored responses. Users will also benefit from increased transparency, as the interface will now display memory sources for generated answers, allowing users to verify, edit, or remove specific data points. While these features are currently rolling out to Plus and Pro users on the web, OpenAI plans to expand access to free, business, and enterprise tiers in the near future.

For the developer community, the new model is now accessible via the API under the designation ‘chat-latest.’ While the legacy GPT-5.3 model remains an option for paid users, it is slated to be phased out within three months. This transition follows OpenAI’s historical pattern of model updates, a process that has previously drawn mixed reactions from users who developed strong personal attachments to the specific conversational styles and personalities of retired models like GPT-4o.