SaaS-Bench

SaaS-Bench is a benchmark comprising 23 deployable Software-as-a-Service systems across six professional domains, containing 106 tasks designed to evaluate Computer-Using Agents in realistic work scenarios that require long-horizon execution and involve both text-only and multimodal settings.

Papers using SaaS-Bench (2)

SaaSBench: Exploring the Boundaries of Coding Agents in Long-Horizon Enterprise SaaS Engineering2026SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS to Solve Professional Workflows?2026