ParentFull threadbobxmax·is Mistral really doing anything here? Llama models are open source, Cohere runs on prem etcView on HN