OpenAI unveils GPT-5.5: AI that increasingly does the work for us

GPT-5.5 represents a significant advance in the field of agent-based artificial intelligence. The model can understand the broader context of a task, plan multiple steps, use different tools, and verify the results itself, rather than requiring the user to control each individual step.

AI as a digital task performer

The biggest leap is in computer work. GPT-5.5 can help with writing and correcting code, researching information, analyzing data, and creating documents and spreadsheets. OpenAI emphasizes that the user can also trust the model with more messy and complex tasks, as it can break them down into steps and bring them to a final result.

Of particular note is the integration with the Codex tool, which is already used by around 4 million developers per week. According to OpenAI, GPT-5.5 can perform at the level of an experienced engineer when programming, as it better understands the structure of systems, identifies errors and predicts the consequences of changes in the code.

The test results confirm the progress. On the Terminal-Bench 2.0 benchmark, the model achieved 82.7 %, which is significantly higher than the previous generation and competing models. Even when solving real programming tasks, the model independently completes more than half of the cases in one attempt.

Expansion into scientific and business work

The progress is not limited to programming, however. GPT-5.5 is also increasingly being used as a tool for scientific research and business processes. The model can help with hypothesis generation, data analysis, and interpretation of results, opening up possibilities for use in a research environment.

The company emphasizes that the model also improves productivity in everyday work. It is already used internally by more than 85 % employees at OpenAI, where they automate analyses, reports, and other complex tasks.

Moving towards the “super app”

GPT-5.5 is also a step towards OpenAI’s broader strategy of merging ChatGPT, Codex, and other AI features into a single platform — a so-called “super app.” The idea is to create a universal tool that can cover most of the digital work of users and businesses.

The company's management emphasizes that this is the beginning of a new way of working with computers, where the user no longer controls every step, but rather defines the goal, and the AI takes care of the implementation.

Faster development and greater competition

The release of GPT-5.5 comes just weeks after the previous version, indicating a remarkable acceleration in development in the AI industry. OpenAI is under pressure from competitors, especially Anthropic and Google, which are also intensively developing advanced models.

At the same time, the company also emphasizes security. GPT-5.5 includes the most stringent protection mechanisms to date, especially in areas such as cybersecurity and biological applications.

GPT-5.5 is not just an upgrade of the model, but a signal of a broader shift: artificial intelligence is becoming an active doer of work.