Anthropic AI passed 41.25%, fell short on freelance tasks at 58.75%

Anthropic’s Fable 5.2 topped Scale AI’s Remote Labor Index with a 41.25% pass rate on remote freelance tasks, while second-place Opus 5.5 reached 21.25%, a 20-percentage-point gap. The result means even the leading model failed 58.75% of tasks against client acceptance criteria. Anthropic had published a 41.3% RLI score for Fable 5.2 at launch, attributing it to a high-effort reasoning setting and safeguards, with the small difference reflecting decimal rounding. The ranking matters because it shows frontier models still fall short on more than half of real-world freelance assignments. Next, watch whether rival models narrow the gap and whether higher-effort configurations translate into meaningfully better acceptance rates.