Same for me, but it produced much better answer than what would I expect from GPT 5.5 xhigh. Also the good think about Codex is that it always finishes the prompt, even when the limit was reached in the half.