Have you used any benchmark to test agents? I'm currently looking for REST API usage benchmark for LLMs.