OpenAI's new coding platform "helped create itself"
An early version of a new coding system from OpenAI was used to debug and write parts of its own code. Human engineers said they were "blown away" by the accelerated development.
OpenAI has unveiled GPT‑5.3‑Codex, a new model that it says merges GPT‑5.2‑Codex’s frontier coding performance with GPT‑5.2’s reasoning and professional-knowledge capabilities and runs 25% faster.
The ChatGPT AI pioneer, in an announcement, said GPT‑5.3‑Codex sets new records on SWE‑Bench Pro and Terminal‑Bench, and shows strong results on OSWorld and GDPval.
The company added that the model was instrumental in "creating itself".
"The Codex team used early versions to debug its own training, manage its own deployment, and diagnose test results and evaluations - our team was blown away by how much Codex was able to accelerate its own development," OpenAI said in the announcement.
"With GPT‑5.3-Codex, Codex goes from an agent that can write and review code to an agent that can do nearly anything developers and professionals can do on a computer."
OpenAI said GPT‑5.3‑Codex can carry out long-running tasks involving research, tool use, and complex execution, and that users can interact with and steer the model while it works in the Codex app. The AI firm noted the model produced autonomous iterations on web games over millions of tokens and improved default outputs for routine web development prompts.
It also reported that teams used Codex to monitor and debug the release’s training runs, optimise the inference harness, identify context rendering bugs and root causes of low cache hit rates, and dynamically scale GPU clusters to stabilise latency during launch.
Related reading
- OpenAI details Codex App Server for agent-based developer tools
- VfL Wolfsburg adopts ChatGPT Enterprise in drive for digital efficiency
- Vikings to use Viper AI to analyse game tape, run draft scouting and review free agents
OpenAI said the model also helped build data pipelines and summarise large evaluation results.
GPT‑5.3‑Codex is, meanwhile, also classified by OpenAI as "High capability" for cybersecurity-related tasks under its Preparedness Framework and said it is the first model they directly trained to identify software vulnerabilities.
The Recap
- OpenAI launched GPT-5.3-Codex, a faster agentic coding model.
- The model achieves state-of-the-art results on multiple benchmarks.
- Available with paid ChatGPT plans; API access coming soon.
