Ⓜ️ Mistral-Large just got released I tried 5 complex prompts to compare Mistral-Large with GPT-4. Here's what I gathered: - Answers are *very* similar, which indicates that Mistral-Large has been trained on a synthetic dataset generated by GPT-4. - It is more concise and inference speed is faster (that's helpful!) - It doesn't have access to tools like Code Interpreter, so it fails some math questions. - A lot less "lazy" in terms of code: it doesn't try to explain what it's going to do and immediately outputs code. It looks pretty on par in general. You can try it for free using the new Le Chat: chat.mistral.ai