Ask HN: Have you guys used the newer Claude models in production?

Hey all!

I have been using the Claude Opus and Haiku models since they dropped and its a very different feeling compared to other LLMs. Haiku with the longer context window and inference speed feels amazing to use and prompt (its not so hard with few-shot examples) Opus just feels much smarter compared to GPT4 at this point.

What has your experience been with these models?