OpenAI has shipped GPT-5.4 and GPT-5.4 mini into Codex, its CLI-based coding agent, alongside a doubling of rate limits across all tiers. The model upgrade brings improved code generation accuracy, better multi-file reasoning, and faster response times. GPT-5.4 mini, the lighter variant, is optimised for quick completions and single-file tasks where speed matters more than depth.
The bigger news for many developers is the Windows release. Codex was previously macOS only, which locked out a significant portion of the developer market. The cross-platform launch brings Codex to Windows and Linux, removing one of the main objections enterprise teams had to adopting it. The desktop app provides the same terminal-based interface with native OS integration for file access and git operations.
Codex usage has already reached 60% of Cursor's volume, a striking figure given that Cursor had a multi-year head start and a more polished IDE experience. The numbers suggest that developers who prefer terminal-based workflows are gravitating towards Codex and Claude Code rather than GUI-based AI IDEs. The CLI-first approach appears to resonate strongly with experienced developers who live in their terminals.
