OpenAI has updated ChatGPT’s default model to GPT-5.5 Instant in a bid to improve accuracy, reduce errors, and deliver more concise responses.
GPT-5.5 Instant is rolling out starting today to all ChatGPT users, replacing GPT-5.3 Instant as the default model and becoming available in the API as chat-latest. For paid users, GPT-5.3 Instant will remain accessible through model configuration settings for three months before it is retired.
OpenAI is also expanding personalisation features for Plus and Pro users on the web. GPT-5.5 Instant can draw on past chats, uploaded files, and connected services such as Gmail to tailor responses. The system determines when personalisation is useful and surfaces relevant context automatically.
The company said these capabilities will soon be extended to mobile devices, with further expansion planned across the Free, Go, Business, and Enterprise tiers in the coming weeks.
The company said GPT-5.5 Instant produces “52.5% fewer hallucinated claims” on high-stakes prompts in areas such as medicine, law, and finance, compared with its predecessor. It also reduced inaccurate claims by 37.3% in conversations that had previously been flagged for factual errors.
“Small improvements make a big difference,” the company said, noting that the model is used daily by hundreds of millions of people. The update focuses on “stronger and tighter answers,” a more natural tone, and improved use of prior context when personalisation is relevant.
The model also shows gains in benchmark evaluations across reasoning and technical domains. On the AIME 2025 math benchmark, accuracy increased to 81.2% from 65.4%. Performance on GPQA, which tests PhD-level science knowledge, rose to 85.6% from 78.5%. Improvements were also reported in multimodal reasoning and document parsing benchmarks.
OpenAI said GPT-5.5 Instant is better at deciding when to use web search and has improved capabilities in analysing images and handling STEM-related queries.
The company also highlighted changes in response style. GPT-5.5 Instant uses fewer words and lines on average while maintaining completeness. “Responses are tighter and more to-the-point without losing substance,” it said.
To address transparency, the company introduced memory sources, which show users what prior context influenced a response. Users can review, delete, or modify stored information, and shared conversations will not expose these sources.
“Users remain in control of what’s in memory,” the company said, adding that temporary chats can be used to avoid storing or referencing past interactions.
OpenAI recently launched GPT-5.5, positioning it as an agentic AI system that can independently execute complex, multi-step tasks across software, data, and research workflows with minimal human supervision.
ALSO READ: The Playground is Closed: 10 Hard Truths from the Cisco AI Summit
