OpenAI Unveils GPT‑5.3‑Codex Model

OpenAI has introduced GPT‑5.3‑Codex, a new model that combines the coding capabilities of GPT‑5.2‑Codex with enhanced reasoning and professional knowledge. The model runs 25% faster than its predecessor, allowing tasks to be completed more efficiently. It can handle long-running tasks that involve research, tool use, and complex execution. Users can monitor its work and provide instructions in real time without losing prior context or results. 

GPT‑5.3‑Codex set new records on SWE-Bench Pro and Terminal-Bench 2.0, excelling in multiple languages and terminal tasks. It can autonomously build complex web apps and games, process millions of tokens, and interpret general or underspecified prompts to produce functional, production-ready websites, including discounted pricing displays and testimonial carousels.

Beyond coding, GPT‑5.3‑Codex supports the full software lifecycle — debugging, deployment, testing, documentation, and data analysis — and assists with professional tasks such as presentations and spreadsheets. It collaborates interactively, responds to feedback, and adapts its actions. The model also has cybersecurity capabilities, identifying software vulnerabilities and supporting defensive research.

Enterprises use GPT‑5.3‑Codex through the Frontier platform, which connects data, provides shared context, and manages access controls. Companies report major productivity gains: reducing production optimization from weeks to one day, increasing client-facing time by over 90%, and raising output by 5%. Frontier lets AI learn from experience, integrate into workflows, and operate within clear rules.