What changed
Anthropic's new model Fable 5.2 completes 41.3 percent of real freelance projects at professional quality, according to the Remote Labor Index (RLI) from Scale Labs and the Center for AI Safety (CAIS). This marks a significant increase from the 2.5 percent achieved by the top system, Manus, a year ago. Fable 5.2 also outperforms its direct predecessor, Fable 5.1, which completed just under 18 percent of tasks.
What this means in practice
The RLI measures how often AI agents complete real freelance jobs well enough for a paying client to accept the work. Fable 5.2 completes more than 16 times as many professional jobs as Manus did a year ago. The model also shows substantial improvement compared to its previous version, Fable 5, which achieved about 16 percent in July. Fable 5.2's performance is roughly two and a half times that of Fable 5. The model's success is attributed to its ability to handle complex tasks such as editing audio and music, designing architecture and interior spaces, and creating 3D models. It also manages long, complex projects that previously top-ranked models regularly failed.
Limitations and availability
While Fable 5.2 performs well in several areas, it still struggles with tasks like industrial product design and animations. When AI agents fail in the test, they typically deliver amateur-quality work, incomplete results, broken files, or contradictory designs. The RLI dataset includes 240 completed projects from 358 verified professionals on Upwork, totaling about $144,000 in value. AI agents have access to a virtual workspace with professional software like Blender, GIMP, or Audacity, along with specialized AI tools for image, audio, and video generation. Human experts manually check whether an AI agent completed a job successfully by comparing the AI's output with the work delivered by human freelancers.
