Fable 5.2 completes 41.3% of real freelance jobs on the Remote Labor Index
On the Remote Labor Index from Scale Labs and the Center for AI Safety, Anthropic’s Fable 5.2 delivered client-acceptable work on 41.3% of real freelance projects, about twice the next-best systems.
Source published · In the Daily Brief of 11 November 2026
Filed under Agents, Large language models
The benchmark uses 240 completed Upwork projects from 358 verified professionals, worth about $144,000 in total. A project costs $632.60 on average and takes a human freelancer just under 29 hours. Trained reviewers open each submission in the relevant professional software and compare it with what the human delivered.
Opus 5.5 scored 21.25% and OpenAI’s GPT-6 Astra 20.83%. When the index launched about a year ago the best agent managed 2.5%, and in July Fable 5 led at about 16%. Scale Labs said the largest gains were in audio editing, architectural design and 3D modelling, while industrial design and animation remain weak.
The Center for AI Safety has so far failed to automate the grading: AI reviewers rated polished but flawed work too generously, according to The Decoder.