Are you guys using Vapi provided LLM your bringing your own? How is the performance of Openrouter OpenAI LLM's?