OpenAI Launches GPT-5.1-Codex-Max: The Most Advanced AI Coding Model for Long-Running Tasks
OpenAI has released GPT-5.1-Codex-Max, its most advanced coding model so far. It is designed to handle long, detailed software projects and stay reliable from start to finish.
OpenAI has released GPT-5.1-Codex-Max, its most advanced coding model so far. It is designed to handle long, detailed software projects and stay reliable from start to finish.
The model is built on GPT-5.1 and trained on real engineering tasks; writing pull requests, reviewing code, building full websites, and answering technical questions. Because of this, it performs much better than previous Codex versions.
One of the biggest upgrades is its “agentic” ability. You can give it a goal, and it can work independently for hours, sometimes over a full day, fixing bugs, improving its own code, and completing the task without constant supervision. It also works smoothly on Windows now, which makes it accessible to far more developers.
OpenAI says its internal engineering team ships around 70% more pull requests after adopting Codex-Max, showing that it’s speeding up real development work, not just producing demos.
Sam Altman praised the team publicly, calling out the rapid progress behind the new model.
GPT-5.1-Codex-Max is available to ChatGPT Plus, Pro, Business, Edu, and Enterprise users. Developers using the Codex CLI will get access once API support is launched. It has also become the default model across all Codex interfaces.
In benchmark tests, Codex-Max shows higher accuracy, fewer mistakes, and lower “thinking token” usage, meaning it works faster and costs less to run. OpenAI has also introduced a deeper reasoning mode for complex tasks where quality matters more than speed.
This launch sets the stage for a growing competition between OpenAI’s Codex-Max and Google’s new Antigravity platform. Coding is quickly moving into an era where advanced AI models work alongside developers as true partners.
