Codex Upgrade Announcement

OpenAI has released GPT-5-Codex, optimized for agentic software engineering. This model is trained to perform real-world software engineering workflows beyond simple code generation, including building projects, adding features, debugging, large-scale refactoring, and code review.

GPT-5-Codex dynamically adjusts its thinking time according to the complexity of the task. It responds quickly to simple requests, while for tasks such as complex refactoring, it can run independently for over 7 hours, fixing test failures and completing the final implementation on its own.

Key features include the following:

• Agent optimization: Better adherence to AGENTS.md instructions, generating high-quality code without the need for separate lengthy style guides.
• Performance verification: Performs all 500 tasks in SWE-bench Verified, demonstrating refactoring ability across various languages including Python, Go, and OCaml.
• Integrated development environment: Enables seamless work across every environment developers use, including CLI, IDE extensions, web, GitHub, and the ChatGPT iOS app.

GPT-5-Codex is included in ChatGPT Plus, Pro, Business, Edu, and Enterprise plans. Developers using the API can access it at the same price as GPT-5 through the Responses API.