Fable 5 doubles remote-work automation
New Remote Labor Index results put Claude Fable 5 at a 16.1% automation rate on 240 real freelance projects, roughly double Opus 4.8 at 8.3%.
The Center for AI Safety and Scale AI Labs published new Remote Labor Index results on July 1. Claude Fable 5 completed 16.1% of real freelance projects at client-acceptable quality, roughly double Claude Opus 4.8 at 8.3%, with GPT-5.5 at 6.3%. The index spans 240 projects across fields like 3D and CAD, architecture, graphic design, video, audio, data analysis, and web apps. Fable 5 was evaluated on 218 of the 240 before its access was restricted in June.
The trajectory matters more than the headline number. When RLI launched, the best system managed 2.5%; the previous published leader was Opus 4.6 with a Claude Cowork scaffold at 4.17%. In less than eight months, the measured ceiling has risen more than fourfold.
The setup is worth knowing before quoting the numbers. Anthropic models ran on Claude Code and OpenAI models on Codex CLI, both with computer use, a worker-critic loop, budgets up to $150 per project for Fable 5, and up to 24 hours of wall-clock time. Human evaluators judged deliverables against professional standards; an automated judge the team tried overestimated the newest models by 2 to 3 times, which is a quiet warning about model-graded evals in general.
Fable 5 itself only returned to general availability on July 1, and its cheaper sibling Sonnet 5 launched the day before. For what these percentages mean in practice, the automation starter playbook is the honest place to start: the 84% that still fails is the reason process beats enthusiasm.