Fable 5.2 posts 41.3 percent on the remote labor index
Nearly 50 of the 240 projects on the Remote Labor Index were solved by no agent except Fable 5.2, which Anthropic announced on November 9. The model finished 41.3 percent of the set, nearly double Opus 5.5's 21.25 percent.
| Fable 5.2 | Opus 5.5 | GPT 6 Astra | |
|---|---|---|---|
| RLI Automation Rate | 41.3% | 21.25% | 20.83% |
GPT 6 Astra sits at 20.83 percent. Gemini Argon 4 at 17.08 percent. A year ago the top score was 2.5 percent.
The index covers 23 of Upwork's 64 domains and excludes physical labor, long-term evaluation, and direct client interaction. Each project averages $632.60 and takes a human freelancer 28.9 hours, with a median of 11.5.
The workspace is virtual, equipped with Blender, GIMP, and Audacity. Fable 5, the family's July entry, managed about 16 percent, and Fable 5.1 just under 18 percent.
The 41 excluded domains and the unquantified failure modes mean the number overstates what a buyer can deploy.
The Elo scores, cost-savings percentages, and domain-level breakdown are not in any readable document.
The constraint for the next evaluation cycle is domain coverage, not model capability.
The draft's lede led with the headline number; this version leads with the unsolved-completion gap, which is the more concrete claim and less generic. The rhythm shifts from a single dense paragraph to short, staggered beats, with a one-sentence paragraph for the missing-data caveat. The title keeps the version and the score, which are the two facts a reader needs to file the piece under.
After @ScaleAILabs. We did not report this. The pictures, if any, are theirs.
