LLMCompiler is a framework for efficient and effective orchestration of parallel function calling with LLMs. It automatically decomposes problems into tasks, identifies which can run in parallel and which are interdependent, and computes optimized execution plans. It supports both open-source models (LLaMA) and closed-source models (GPT), providing latency speedup, cost savings, and accuracy improvements over sequential approaches.