OpenAI Again Discontinues Dedicated Coding Model Codex, Integrating It Into GPT-5.5

Following the release of GPT-5.4, OpenAI has fully merged its standalone Codex programming model into its primary architecture. Romain Huet, Head of Developer Experience at OpenAI, confirmed that the company has phased out its dedicated coding model division, making the GPT-5.3 release in early February the final standalone Codex version. Huet highlighted that GPT-5.5 introduces significant advancements in autonomous coding, allowing AI to independently manage complex development workflows. The update also delivers enhanced computer-use capabilities and stronger performance across general-purpose tasks.

Regarding computational efficiency, Huet revealed that GPT-5.5 requires fewer tokens than GPT-5.4 to complete identical programming assignments, delivering higher-quality results with reduced resource consumption. Nevertheless, despite these efficiency gains, OpenAI will implement an approximate 20% price increase for its API services.

This strategic shift mirrors a recurring pattern in OpenAI’s development history. The tech giant originally discontinued its dedicated Codex model in 2023 to focus on universal large language models. Codex returned in May 2025 as Codex-1, powered by the o3 architecture, alongside the Codex AI agent platform. Currently under active development, the AI agent software has been elevated to a core strategic initiative, positioned alongside ChatGPT as a primary focus for the company.