Beyond GPT-5: Making LLMs Cheaper and Better via Performance–Efficiency Optimized Routing - Arxiv Papers | Wave AI Podcast Notes