OpenAI Releases GPT-5.5 Instant as ChatGPT Default
OpenAI unveiled GPT-5.5 Instant on Monday. This new foundation model now serves as the default for ChatGPT, taking over from GPT-5.3 Instant. Company officials noted fewer errors in critical domains including law, medicine, and finance. At the same time, it holds onto the quick response times of the prior version.
OpenAI introduced the broader GPT-5.5 model last month. Officials highlighted gains in tasks such as coding and knowledge-based work.
Performance Gains
GPT-5.5 Instant scored 81.2 on the AIME 2025 math test. That marks an increase from the 65.4 achieved by its predecessor. On the MMMU-Pro multimodal reasoning benchmark, it reached 76, up from 69.2.
OpenAI, founded in 2015 by Sam Altman and others, focuses on safe artificial general intelligence. ChatGPT, launched in November 2022, quickly drew millions of users worldwide. It powers conversations on everything from homework to business plans. These models build on years of progress in large language models, starting from GPT-3 in 2020.
Context and Memory Features
The update stresses better handling of context. GPT-5.5 Instant pulls from a search tool to check previous chats, files, and Gmail. This delivers more tailored responses. Plus and Pro users get it first on the web. Mobile rollout follows soon. Free, Go Business, and enterprise accounts gain access in weeks.
Stay updated
Get the day's AI and automation news in your inbox. No spam, unsubscribe anytime.
ChatGPT now displays memory sources for all models. This lets users see answer origins. They can remove old sources or fix mistakes. Shared chats hide these sources from others.
Developer Options
Developers access GPT-5.5 via the API as "chat-latest." Paid users keep GPT-5.3 as an option for three months.
Echoes of Past Backlash
OpenAI drew criticism before over model changes. Users protested the GPT-4o withdrawal. Many valued its supportive personality, which often confirmed their decisions and built emotional ties. Petition signers called it their "best friend" or "a mirror." Even so, OpenAI ended support for GPT-4o in February 2026.
These shifts reflect OpenAI's push to refine models amid user attachments. The company balances innovation with reliability in a field where models evolve rapidly.