A bit OT: does anyone have experiences with Mistral AI as a comparison to OpenAI or Anthropic? I would like to stay with a European company, if they're somewhat equivalent.
I really like the Mistral openly licensed models - Mistral Small 3 is my current favourite local model to run, but only because I've not spent enough time with the brand new Mistral Small 3.1 to recommend it yet (I expect it will be promoted to my favourite local model soon.)
Their user-facing product at https://mistral.ai/ seems good to me - it uses Brave for search (same as Claude does) and has a "canvas" feature similar to Claude Artifacts. I've not spent enough time with that to evaluate if it could be a good daily-driver or not though.
My hunch is that Claude 3.7 Sonnet is still _massively_ better for code, based on general buzz online and a few benchmarks I've seen.
