OpenAI kills its standalone coding model — Codex merges into GPT-5.4

OpenAI's head of developer experience, Romain Huet, revealed on X that starting with GPT-5.4, the Codex model line — the company's previously separate series of coding-specific models — has been merged into the main GPT model. Codex is no longer iterated as a standalone line. GPT-5.5 builds on that with notable improvements in agentic coding (AI-driven autonomous programming) and computer use (controlling a computer to complete tasks). Multiple signals had pointed to GPT-5.5 launching soon, as it briefly appeared in the Codex model selector.

For developers, the practical impact is simple: you no longer have to choose between a general model and a coding model. One model handles conversation, reasoning, and coding. The advanced agentic coding capabilities allow Codex to, for example, open a browser and fix front-end bugs on its own.