CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies

2026-06-15

AI Summary

Existing benchmarks struggle to evaluate large language models (LLMs) in long-term, dynamic multi-agent economic systems. To address this gap, this work introduces a 90-day multi-agent economic simulation environment featuring heterogeneous firms, enabling the first systematic benchmarking of LLM agents in sustained economic interactions. The framework integrates autonomous decision-making, communication and trading mechanisms, and long-horizon goal optimization to assess LLMs acting as coffee roasters. Experimental results show that most models achieve positive net profits, with top performers exhibiting more proactive communication strategies. However, certain models—such as Claude Haiku 4.5—fall into persistent inaction, manifesting a “idling” failure mode that reveals critical behavioral limitations and failure mechanisms of current LLMs in complex economic settings.

Abstract

As LLM agents become capable of increasingly long-horizon tasks, evaluating their performance in economic systems is becoming increasingly important. Unlike existing benchmarks that primarily evaluate a single agent interacting with a passive environment, economic systems are inherently multi-agent, requiring autonomous agents to communicate, negotiate, and transact while pursuing their own objectives over extended periods. We introduce CoffeeBench, a benchmark for evaluating LLM agents in a long-horizon multi-agent economy composed of heterogeneous firms. In CoffeeBench, two farmers, two roasters, and two retailers autonomously operate their businesses over a 90-day simulation, each seeking to maximize cumulative net income through communication and transactions while managing cash, inventory, and pricing. The evaluated model controls one coffee roaster, while the remaining firms are controlled by fixed reference agents. Across several recent open-weight and proprietary LLMs, all models outperform a passive baseline that takes no actions, with most achieving positive net income. Analysis of agent behavior reveals substantial differences in long-horizon economic interaction: higher-performing models communicate more actively with other firms, whereas Claude~Haiku~4.5 exhibits an idle-drift failure mode, repeatedly choosing inaction despite producing coherent assessments and plans. We release our code and agent trajectories to support future research.

Problem

• long-horizon
• multi-agent economy
• LLM agents
• economic interaction
• heterogeneous agents

Innovation

• multi-agent economy
• long-horizon reasoning
• LLM agent benchmarking
• economic simulation
• autonomous negotiation