GPT-5.4

GPT-5.4 is OpenAI's most powerful and efficient frontier model, released simultaneously in ChatGPT, the API, and Codex. It brings together advances in reasoning, coding, professional knowledge work, and agentic computer use into a unified model.

**Three variants for every use case**: - **GPT-5.4** (Standard) — Available via API and Codex; the best general-purpose model for developers building agentic workflows - **GPT-5.4 Thinking** — The ChatGPT-facing variant; shows an upfront plan of its thinking process, lets users course-correct mid-response for tighter alignment with intent - **GPT-5.4 Pro** — Maximum performance for complex, high-stakes tasks; available in ChatGPT Pro and via API

**Headline capabilities**: - **1M token context** — Plan, execute, and verify long-horizon tasks across massive codebases or document sets - **Native computer use** — First general-purpose OpenAI model with state-of-the-art computer operation capabilities (OSWorld-Verified: 75.0%); agents can open applications, navigate UIs, and verify their own work - **Tool Search** — Intelligently selects the right tool from large ecosystems of connectors, dramatically improving multi-tool agentic reliability (Toolathlon: 54.6%) - **Token efficiency** — Significantly fewer reasoning tokens vs GPT-5.2, translating to lower API costs and faster responses - **Professional knowledge work** — GDPval score of 83.0% (matches or exceeds industry professionals in 83% of tasks across 44 occupations) - **Coding excellence** — SWE-Bench Pro (Public): 57.7%, inheriting GPT-5.3-Codex's industry-leading coding capabilities

GPT-5.4 is available to all ChatGPT users (Thinking variant), developers via the OpenAI API, and Codex users. Pricing follows OpenAI's standard token-based API rates.