ChatGPT gets new GPT-5.5 Instant model

OpenAI has released a new default model called GPT-5.5 Instant, which will replace GPT-5.3 Instant as the main model powering ChatGPT.

According to the company, the updated model reduces hallucinations in sensitive areas such as law, medicine, and finance, while maintaining low latency.

Last month, OpenAI introduced the broader GPT-5.5 model family, highlighting improvements in coding and knowledge retrieval. In the AIME 2025 mathematics benchmark, GPT-5.5 Instant scored 81.2 points compared to 65.4 for the previous version. The model also improved its results in the multimodal MMMU-Pro benchmark, reaching 76 points versus 69.2 previously.

A major focus of the update is context management. GPT-5.5 Instant can use search functionality to reference previous conversations, uploaded files, and Gmail messages to provide more personalized responses. The feature is currently available to Plus and Pro users on the web version of ChatGPT, with mobile support expected soon. OpenAI says it plans to expand access to free users, as well as Go, Business, and Enterprise customers, in the coming weeks.

With the update, ChatGPT will also begin showing memory sources across all models, allowing users to see where certain responses originate. Users will be able to remove outdated memory entries or correct them if responses are inaccurate. Shared chats, however, will not expose memory sources to other users.

For developers, GPT-5.5 will be available through the API under the name “chat-latest.” GPT-5.3 will remain available as an optional model for paid users for the next three months.