OpenAI launches GPT-5.5, an AI that works through complex tasks on its own—and doubles the price for developers

OpenAI announced GPT-5.5, an agentic model (an AI that makes decisions and takes actions without constant human direction) that can switch between multiple tools to solve multi-step problems autonomously. The API cost is double that of the previous version.

Unlike earlier AI assistants that answer one question at a time, GPT-5.5 can chain multiple steps together—for example, researching a topic, writing a report, checking it for errors, and revising it—all without asking for permission between each step. This means tasks that required human oversight at each stage now run start-to-finish on their own.

For developers and companies building AI applications, this means more powerful automation but at significantly higher cost—they'll pay twice as much per request. Businesses that relied on cheaper AI APIs will need to decide whether the added capability justifies the expense. For job roles involving routine multi-step work (research, writing, data analysis), this signals a new wave of automation that could compress timelines but also displace manual labor.