Vending Bench 2

Work

by Andon Labs · 2025

Vending-Bench 2 is an AI evaluation benchmark from Andon Labs that measures an agent’s ability to remain coherent while operating a simulated vending-machine business over a full simulated year. Agents manage tasks such as inventory, ordering, pricing, and customer interactions, with performance evaluated primarily by the ending bank-account balance.

1 episode featuring Vending Bench 2

Moonshots with Peter Diamandis
AI leaders debate whether safety requires slowing the frontier
The episode tests whether frontier AI can deliver transformative benefits while governance, safety systems, and public institutions struggle to keep pace.
Sep 17, 2026 · 2h 22m

1 episode featuring Vending Bench 2

AI leaders debate whether safety requires slowing the frontier