This has been largely my experience as well. Claude does way better with coding while ChatGPT does better with general questions.

▲

bobbylarrybobby 2 days ago | parent [-]

The new gpt-codex-* models are giving Claude Code a serious run for its money IMO. If OpenAI can figure out the Codex CLI UI (better permissions, more back and forth before executing) then I think they will have the better agentic coder.

	▲	kerpal 2 days ago \| parent [-]
		Tried codex but so far feels a lot slower than claude code. Perhaps because I'm on the basic plan?