For all the test cases we have, it'd be instructive to run them through leading language models trained on code.

https://huggingface.co/spaces/bigcode/bigcode-models-leaderboard