OpenAI's new GPT-5.3-Codex is 25% faster and goes way beyond coding now - what's new

OpenAI announced GPT-5.3-Codex, a faster (25%) and more capable agentic coding model that can be steered mid-task, handle long-running processes, and reportedly helped debug and deploy parts of itself. The release emphasizes broader workflow support beyond coding (e.g., debugging, deployment, documentation, and UI work), improved handling of underspecified prompts, and strong benchmark results (SWE-Bench Pro, Terminal Bench, OSWorld, GDPVal). OpenAI also highlights cybersecurity measures—classifying the model as high capability for cyber tasks, deploying an expanded safety stack, launching Trusted Access for Cyber, and offering $10M in API credits for security research. GPT-5.3-Codex is available to paid ChatGPT users via the Codex app, CLI, IDE extensions, and web, with API access planned.