FrontierHarness v1.1: the model sets 92.8% of agent cost
AI-generated summary Runta published FrontierHarness Eval v1.1 on Nov. 17, evaluating 18 coding-agent harnesses across 27,280 runs and 83B tokens. Runta reported that the model explains 92.8% of agent cost variance, while the harness accounts for 2.3%.