Nous Research Released DeepHermes 3 Preview: A Llama-3-8B Based Model Combining Deep Reasoning, Advanced Function Calling, and Seamless Conversational Intelligence