The Remote Labor Index (RLI) is a benchmark of 240 long-horizon remote labor tasks, collected as a representative sample of fully-remote labor performed by freelancers. Tasks in the dataset take human professionals between several hours to several hundred hours (see figures from the paper below).

Though most of the benchmark is private, the creators will occasionally report model performance on the benchmark as evaluated by themselves and/or model providers. At the end of each listed year, I will resolve to the highest reported RLI automation rate (fraction of tasks completed) of an AI model that year. The current best evaluated model as of October 2025 is Manus, at a 2.5% automation rate.