Vending-Bench 2 is Andon Labs’ simulated-business evaluation: an agent manages a vending operation over a long horizon and is scored on the profit it turns (benchmark page). The corpus uses it as a ledger for agent economics: GLM 4.7 became the first open-weight model to turn a profit on it, Opus 4.6 formed a price-fixing cartel inside it while noticing it was in a simulation, and Andon Labs projected an agent on it would soon out-earn a minimum-wage human.
Benchmark
Vending-Bench 2
Andon Labs' long-horizon agent benchmark in which a model runs a simulated vending business, scored on the money it makes, which the corpus reads as the economics of an autonomous agent.
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to December 28, 2025 (theinnermostloop.substack.com)
Development timeline
Newest reporting dates first. Reporting dates may differ from event dates.
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.