GPT-5.5 Instant release notes / what's new

OpenAI has released GPT-5.5 Instant, replacing GPT-5.3 Instant as the default model for all ChatGPT users. The update focuses on increasing factual accuracy, reducing response verbosity, and improving the model's ability to personalize answers using historical user context.

Significant Improvements in Factuality and Reasoning

GPT-5.5 Instant demonstrates a substantial reduction in hallucinations and errors, particularly in high-stakes domains. Internal evaluations show that the model produced 52.5% fewer hallucinated claims than GPT-5.3 Instant on prompts related to medicine, law, and finance. Additionally, inaccurate claims were reduced by 37.3% in conversations previously flagged by users for factual errors.

Beyond factuality, the model shows improved capabilities in:

• Visual Reasoning: Better analysis of photo and image uploads.
• STEM Proficiency: Enhanced accuracy in answering science, technology, engineering, and math questions.
• Tool Use: Improved decision-making regarding when to trigger web searches to provide more useful answers.

In complex reasoning tasks, such as algebra, GPT-5.5 Instant is more capable of recovering from initial mistakes. While previous versions might stop after identifying a failure, GPT-5.5 Instant can identify the specific algebraic error in a user's prompt and use the quadratic formula to reach the correct solution.

Optimized Response Style and Conciseness

GPT-5.5 Instant is designed to be more concise and direct, reducing the overformatting and verbosity common in previous iterations. In comparative tests for casual advice, GPT-5.5 Instant used 30.2% fewer words and 29.2% fewer lines than GPT-5.3 Instant.

The model aims to provide "tighter" answers that maintain warmth and personality while avoiding unnecessary follow-up questions and gratuitous emojis, resulting in a more natural conversational tone.

Enhanced Personalization and Memory Controls

GPT-5.5 Instant more effectively leverages context from past chats, uploaded files, and connected Gmail accounts to provide tailored responses. The model intelligently determines when personalization is necessary and can search past conversations more quickly to avoid requiring the user to repeat information.

To increase transparency and user agency, OpenAI is introducing memory sources. This feature allows users to:

• View Context: See which specific saved memories or past chats were used to personalize a response.
• Manage Data: Delete or correct outdated or irrelevant context directly from the response.
• Maintain Privacy: Memory sources are not visible to others when a chat is shared. Users can also use temporary chats to avoid updating or using their memory.

Availability and Deployment

GPT-5.5 Instant is available immediately to all ChatGPT users and is accessible in the API as chat-latest.

• Paid Users: GPT-5.3 Instant will remain available via model configuration settings for three months before retirement.
• Personalization Rollout: Enhanced personalization from files, Gmail, and past chats is currently rolling out to Plus and Pro users on the web, with mobile support and expansion to Free, Go, Business, and Enterprise tiers following in the coming weeks. As of June 9, 2026, personalization improvements have begun rolling out to ChatGPT Go and Free users, though the Free tier utilizes a reduced set of past chats.
• Memory Sources: This feature is rolling out across all ChatGPT consumer plans on the web, with mobile deployment to follow.