As I’m sure many of you already know, the European AI company Mistral (best known for the open-source LLM model Mistral-7B) just released Mixtral-8x7B which is a mixture of experts model that has jostled the open-source community by actually being close to comparable to ChatGPT. Along with this, on their website they have also annouced mistral-medium, which seems to be an API-only model a step above their MoE model. Has anyone had access to it or been able to compare it to good ol’ GPT-3.5 Tur...