I’ve mostly switched to Codex (GPT-5.4 high) over Claude Code (Opus 4.6) in the last few weeks. I think my style of workflow and prompting seems to generate a bit better results with it. I have a thin CLI based issue tracker and an associated skill that I use instead of markdown files and Codex does a better job of interacting with it and updating existing tasks over time whereas Claude Code tends to create duplicate tasks with slight changes.
I find the code quality of both to be pretty on par with some style differences; Claude has been leaving very verbose comments as of late
I do slightly prefer the Codex app over the Claude Code TUI and Visual Studio Code integrations.
I do slightly prefer the Codex app over the Claude Code TUI and Visual Studio Code integrations.