Comment by pxc

So far, I like the answer quality and its voice (a bit less obsequious than either ChatGPT or DeepSeek, more direct), but it seems to badly mangle the format of its answers more often than I've seen with SOTA models (I'd include DeepSeek in that category, or close enough).