CostBench is a newly introduced benchmark designed to evaluate Large Language Model (LLM) agents on their ability to generate and adapt cost-effective plans within dynamic environments. Unlike existing assessments that primarily focus on whether tasks are completed, CostBench emphasizes the importance of resource efficiency and adaptability in multi-turn planning scenarios. This benchmark addresses a notable gap in current evaluation methods by prioritizing cost-optimal planning, which is critical for practical applications where resource constraints and changing conditions are common. By focusing on these aspects, CostBench aims to provide a more comprehensive measure of an LLM agent's performance in real-world settings. The development of this benchmark reflects ongoing efforts in the AI community to enhance the practical utility of LLM tool-use agents beyond mere task completion. CostBench’s introduction is supported by recent related research that similarly highlights the need for evaluating adaptability and efficiency in AI planning. Overall, CostBench represents a significant step toward more nuanced and realistic evaluation frameworks for AI agents operating in complex, evolving environments.