OpenAI has added a new development to its recent advancements by introducing the GPT-5.5 Instant model. This model will replace the previous version, GPT-5.3 Instant, as the default model for ChatGPT.
New Features and Performance Improvement
One of the most notable aspects of the new model is its reduction in hallucination rates in legal, medical, and financial fields. According to OpenAI's statements, this model aims to provide more accurate results while offering lower latency.
Additionally, GPT-5.5 Instant achieved a significant success by scoring 81.2 points on the AIME 2025 math test, surpassing the previous version's score of 65.4 points. It also demonstrated superior performance in multimodal reasoning tests, scoring 76 points.
According to the company's data, the new model produces 52.5% less erroneous information and reduces the number of flagged inaccuracies by 37.3%.
Personalized Responses and Memory Resources Feature
GPT-5.5 Instant provides more concise answers by avoiding unnecessary emojis and can access past conversations through its search function. This feature will initially be active for Plus and Pro users, with plans to roll it out to mobile platforms soon.
Moreover, with the “memory resources” feature developed for all ChatGPT models, users will be able to see the context in which personalized responses were generated. Users can edit or delete this information if necessary.
OpenAI also stated that memory resources will not be visible when shared with others. Currently, this feature is available for ChatGPT consumer plans on the web.
In addition to all this, the GPT-5.5 model will be made available to developers via API as “chat-latest” and has started to meet all users; GPT-5.3 Instant will remain available for three months until it is phased out.