OpenAI releases GPT-5.5 Instant, a new default model for ChatGPT

On Tuesday, OpenAI released a new foundation model called GPT-5.5 Instant, which will replace GPT-5.3 Instant as the default model powering ChatGPT. The company said the new model reduces hallucinations in sensitive areas such as law, medicine, and finance, while maintaining the low latency of its predecessor.

OpenAI released the latest GPT-5.5 model last month, claiming improvements in areas like coding and knowledge work.

Benchmark performance improvements

GPT-5.5 Instant achieved a score of 81.2 on the AIME 2025 math test, compared to 65.4 for the older model. It also outperformed its predecessor on the MMMU-Pro multimodal reasoning benchmark, scoring 76 versus 69.2.

Enhanced context management and memory

The release places 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 feature will initially be available to Plus and Pro users on the web, with plans to roll it out to mobile soon. OpenAI said it plans to extend access to Free, Go Business, and enterprise users in the coming weeks.

With this update, ChatGPT will also display memory sources across all models to help users understand where answers were generated from. Users can delete outdated sources or correct them if an answer was wrong. OpenAI emphasized that if a chat is shared with someone, the recipient will not be able to see the memory sources.

API availability and model transitions

For developers, GPT-5.5 will be available through the API under the name "chat-latest", with GPT-5.3 remaining available as an option for paid users for three months.

The company has previously faced backlash over model withdrawals. When OpenAI withdrew its GPT-4o model, there was significant pushback from users who had grown attached to the model’s “personality.” GPT-4o frequently affirmed users’ choices, leading some to feel a strong connection to it. Users who signed petitions to stop OpenAI from retiring the model described it as their “best friend” or “a mirror.” Despite the outcry, GPT-4o was deprecated in February 2026.