On April 23, OpenAI officially introduced the new GPT-5.5 model — the next update to the GPT-5 line, which is already being called one of the company’s most important updates in recent times.
On April 23, OpenAI officially introduced the new GPT-5.5 model — the next update to the GPT-5 line, which is already being called one of the company’s most important updates in recent times.
The release was simultaneously reported by leading Western media outlets, including The Verge and Axios, as well as the company itself on its blog.
GPT-5.5 is OpenAI’s new «frontier» model, which the company says works better with:
OpenAI emphasizes that the model has become more «agentic» — that is, it better plans actions, uses tools, and completes tasks.
According to The Verge, GPT-5.5 significantly improves code handling, research, and performance, and also handles multi-step tasks and error checking better.
According to OpenAI co-founder Greg Brockman, the new model is another step towards creating a universal AI product.
In particular, as TechCrunch writes, the company is considering GPT-5.5 as part of a strategy to build the so-called «AI-super app» — a single interface for work, research, and task automation.
Axios adds that the model is focused on «agentic computing» — an approach where AI performs complex tasks almost independently.
According to OpenAI’s official description, GPT-5.5:
In tests, the model shows significant progress in generic tasks, financial modeling, document work, and scientific research.
It is separately noted that GPT-5.5 can act as a «co-developer» or even a «co-researcher», helping to go through the full cycle — from idea to result.
The release is happening gradually. The model is already available in ChatGPT for Plus, Pro, Business, Enterprise plan subscribers, and is also integrated into Codex and developer tools.
A separate version of GPT-5.5 Pro is being launched for more complex tasks.
The release of GPT-5.5 comes against the backdrop of sharply intensifying competition in the AI market.
In particular, OpenAI recently introduced a specialized model for cybersecurity, GPT-5.4-Cyber, Anthropic is actively promoting its own models for enterprise, and major players are betting on «agent» systems.
As Axios notes, new models are increasingly focused not on answers but on completing tasks — from code to research.
