Anthropic’s latest model, Fable 5.2, topped Scale’s Remote Labor Index (RLI) on November 12, 2026, with a 41.3% pass rate for complete freelance deliverables, according to CryptoSlate. The benchmark measures AI‑generated work against professional reference outputs across a diverse set of 240 projects, ranging from documents and audio to 3D models and CAD files. The next‑best performer, Opus 5.5, achieved a 21.25% pass rate, placing it well behind Anthropic’s offering.

How the Remote Labor Index Works

The RLI, developed by researchers at the Center for AI Safety and Scale AI, requires AI systems to produce finished products that meet or exceed a professional reference standard. Human reviewers evaluate each deliverable against the original brief, using majority voting to determine whether the AI’s output satisfies the client’s expectations. Projects that involve direct client interaction, teamwork, or physical labor were excluded from the study, limiting the benchmark’s scope to tasks that can be objectively inspected.

Anthropic’s Results and Caveats

Anthropic labeled its 41.3% score as “High effort” in a November 9 release, noting that safeguards were enabled during testing. However, the company did not disclose spending per project or the extent of human assistance involved. The RLI methodology still leaves a majority of evaluated projects below the acceptance threshold, meaning that while Fable 5.2 outperformed peers, most freelance tasks remain unsolved by the model.

What to Watch Next

Future releases from Anthropic and competing firms will likely aim to improve pass rates and expand the range of task categories. Observers will watch for any disclosure of cost structures or human‑in‑the‑loop support, as these factors are critical for assessing the commercial viability of AI‑assisted freelance work.