Released in early March 2026, this model is OpenAI's first general-purpose system to fold the Codex line into the mainline GPT family, combining stronger reasoning, agentic coding, and tool execution in a single endpoint. The official launch describes it as the most capable and efficient frontier model for professional work, explicitly inheriting the coding strengths of GPT-5.3-Codex while improving how it handles spreadsheets, presentations, documents, and multi-step tool workflows. A higher-tier Pro variant was released alongside the standard model for users who want maximum accuracy on the hardest tasks.

It is the first general-purpose OpenAI model shipped with native computer-use capabilities, letting agents operate desktop and browser interfaces end to end rather than relying on bolt-on tools. Developer guides document five reasoning effort levels (none, low, medium, high, xhigh) that let callers trade depth for cost per request, with the top setting reported at roughly three to five times the cost of the lowest. The model accepts text and image inputs, outputs text only, and pairs a the cataloged API limit input window with a the cataloged API limit output ceiling for a combined context just over one million tokens. Practical fit centers on long-horizon coding agents, deep research, complex document analysis, and multi-application automation, though OpenAI's April 2026 release of the GPT-5.5 successor means it is no longer the newest frontier offering.