SaaSBench: Exploring the Boundaries of Coding Agents in Long-Horizon Enterprise SaaS Engineering

SaaSBench introduces a coding-agent benchmark targeting full-stack enterprise SaaS complexity, including heterogeneous environments and multi-service orchestration, gaps absent from existing benchmarks like SWE-bench. It addresses the structural simplicity problem in current long-horizon coding evaluations. Practitioners building or evaluating agents for real enterprise deployments have a more realistic stress-test available.