OpenAI has released GPT-5.5 Instant, a new default model for ChatGPT, which replaces GPT-5.3 Instant. The GPT-5.5 Instant model reduces hallucination in sensitive areas such as law, medicine, and finance, while maintaining low latency.
Overview
The new model improves upon its predecessor in several areas, including coding and knowledge work. It achieved a score of 81.2 in the AIME 2025 math test, compared to 65.4 for the older model, and outperformed its predecessor on the MMMU-Pro multimodal reasoning benchmark, with a score of 76.0 vs 69.2.
What it does
GPT-5.5 Instant can use its search tool to refer back to past conversations, files, and Gmail to give more personalized answers. This feature will be available to Plus and Pro users on the web, with plans to roll it out to mobile soon. OpenAI plans to extend access to this feature to Free, Go Business, and enterprise users in the coming weeks. The model also allows users to delete outdated sources or correct them if the answer was wrong, and users can see where the answers were generated from.
Tradeoffs
The release of GPT-5.5 Instant means that GPT-5.3 will only be available as an option for paid users for three months. The company has faced backlash from users in the past when withdrawing previous models, such as GPT-4o, which was deprecated in February 2026 despite user petitions to keep it.
In practical terms, the update to GPT-5.5 Instant will improve the overall user experience of ChatGPT, allowing for more nuanced and multi-turn conversations. Developers will also have access to the GPT-5.5 model through API as ‘chat-latest’. As with any update, there may be some adjustments needed for users who are accustomed to the previous model, but the new features and improvements should provide a better experience overall.