About
AgentRealTaskBench (ARTB) is an open, human-run benchmark for personal AI agents and assistants.
Each task is something a real person would ask. Each score links to the recording. The whole task set lives in a public repo.
ARTB is run by David Li, founder of SandBase, which builds infrastructure for AI agents. To keep it neutral: SandBase and directly competing infrastructure products are not evaluated; agents are tested as shipped, with no tools or plugins added by us; no sponsorships, no paid placement, and we pay for every subscription we test; vendors can check facts 48 hours before publication, and corrections are logged in public.
Think a score is wrong, or shipped a fix since we tested? Open task feedback and we will retest.