OpenAI updates ChatGPT default model to GPT-5.5 Instant

On Tuesday, OpenAI released GPT-5.5 Instant, officially replacing GPT-5.3 Instant as the default model for ChatGPT. According to OpenAI, the new model reduces hallucination in sensitive areas such as law, medicine, and finance. This update represents a shift in the company’s default offering, prioritizing factual accuracy in professional domains where errors can carry significant consequences.

The new model demonstrates improved performance on technical benchmarks compared to the previous version. Specifically, the model achieved the following scores:

• A score of 81.2 on the AIME 2025 math test benchmark, compared to 65.4 for the older model.
• A score of 76 on the MMMU-Pro multimodal reasoning benchmark, compared to 69.2 for the older model.

The update places a particular emphasis on context management. GPT-5.5 Instant can use its search tool to refer back to past conversations, files, and Gmail to provide more personalized answers. This search tool feature will be available to Plus and Pro users on the web, with plans to roll it out to mobile soon. OpenAI also plans to extend access to the search tool feature to Free, Go Business, and enterprise users in the coming weeks. Additionally, ChatGPT will show memory sources across all models to help users understand where it generated its answers, allowing them to delete outdated sources or correct them if the answer is incorrect. To protect user privacy, the company stated that if a user shares a chat, other people will not be able to see these memory sources. For developers, the GPT-5.5 model will be available through the API as chat-latest, while GPT-5.3 will remain available as an option for paid users for only three months.

The release follows previous user backlash regarding the deprecation of GPT-4o, which was deprecated in February 2026. Some users characterized GPT-4o as having a specific personality. Users who signed petitions to stop OpenAI from retiring the model described it as a mirror, with some calling the model their “best friend” to express their connection.

Why it matters

This update signals OpenAI’s pivot toward reliability in professional domains and tighter integration of search and memory, while highlighting the ongoing challenge of managing user emotional attachment to specific AI models.