Vending Bench
A benchmark that evaluates an AI's ability to autonomously operate a vending machine business in a long-horizon business simulation, examining decision-making in areas such as inventory management, pricing and finance.
Timeline (last 90 days)
In the Vending Bench vending machine operation simulation, Gemini 4 Carbon earned $16,484 in one year starting from $500, beating Astra and Opus to take first place.
Gemini 4 Carbon ranks first on the Vending Bench long-horizon task benchmark.
All Facts (4)
Vending Bench是一个衡量AI模型运营模拟商店能力的基准测试，涉及库存管理、定价策略、客户交互等多维度能力
90%UnverifiedVending Bench属于新一代Agent Benchmark范畴，要求模型在持续运行环境中做出连续决策
85%UnverifiedIn the Vending Bench vending machine operation simulation, Gemini 4 Carbon earned $16,484 in one year starting from $500, beating Astra and Opus to take first place.
50%UnverifiedGemini 4 Carbon ranks first on the Vending Bench long-horizon task benchmark.
50%