According to the company's statement, the new model understands users' goals faster than before and can take responsibility for entire work processes.
GPT-5.5 excels especially at writing and debugging programming code, research based on information retrieval, and moving between different pieces of software and tools.
A user can give the model a complex and even unstructured task, and the AI will be able to plan the work steps itself, use the necessary tools, and navigate through the different stages until the task is completed.
Although the model's intelligence has risen to a new level, OpenAI says its speed matches that of the GPT-5.4 version.
However, the model is more efficient, as it uses fewer tokens to perform the same complex tasks. Particular attention has been paid to the model's ability to use a computer like a human, enabling it to interpret information displayed on the screen, perform mouse clicks, and type text across different applications.
According to early testers, GPT-5.5 demonstrates exceptional conceptual clarity and can manage large, complex wholes in a way that previous AI models have not been able to.
With the new release, OpenAI has introduced its strongest safety measures to date, aimed at preventing misuse of the model, especially in matters related to cybersecurity and biology.
The company has developed stricter filters to detect high-risk requests, even though this may initially appear to users as the model refusing queries more readily than earlier versions.
GPT-5.5 becomes immediately available to Plus, Pro, and enterprise subscribers in the ChatGPT and Codex services. For API developers, the model is coming soon, and its pricing has been set at 5 dollars per million input tokens and 30 dollars per million output tokens.
An even more powerful GPT-5.5 Pro version will also be offered for more demanding expert tasks.
Written by: Janne Yli-Korhonen @ 24 Apr 2026 5:06
