Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
Terminal-Benchtbench.ai · Low evidenceTerminal Bench is a collection of Harbor-native benchmarks for evaluating AI agents in terminal environments. It provides tasks like building a Linux kernel, configuring a Git webserver, and training a FastText model to measure agent performance.
Parse Score