GPT 5.6 Sol Still Fails on Basic Code Review

chatgpt.com

1 pointmbeavitt4 comments

I'm not going to write some big blog post here, but this chat log is quite entertaining. I wrote a fundamental string algorithm in C and GPT 5.6 Sol on medium effort hallucinated not one but two correctness issues with the code!

Maybe this is not surprising to others but it's been a long while since I caught one of these models making such a glaring error.

I'm pretty sure there is a completely different model for Chat vs Coding.

Like if you deploy 5.6 on azure foundry there is a chat model and a coding/reasoning model

I learned something new today

And maybe it's just the harness and the backing model is mostly the same. But I would expect Codex to give better responses to code questions than chatGpt using the "same" model.

I went back through Azure foundry, and it looks like they have gotten rid of reasoning variants, so might just be one model now.

Reply on news.ycombinator.com