This page may contain stale information. Last updated: 2026-07-02
Definition
The Remote Labor Index (RLI) is a benchmark jointly developed by the center-for-ai-safety and Scale Labs measuring how often AI agents complete real, economically valuable freelance projects at professional quality acceptable to a paying client.
Methodology
- 240 projects worth $144,000 combined, sourced from 358 verified freelancers
- Sectors: 3D/CAD, architecture, graphic design, video/animation, audio, data analysis, web apps
- Human evaluators score deliverables against gold-standard professional work
- Automation rate = share of projects judged equal to or better than human professional standard
July 2026 Leaderboard
| Model | Automation Rate |
|---|---|
| claude-fable-5 Fable 5 | 16.1%* |
| claude-opus-4-8 Opus 4.8 | 8.3% |
| gpt-56 GPT-5.5 | 6.3% |
*Fable 5: 218/240 projects evaluated before U.S. government access restriction. Worst-case (all 22 missing = fail): 14.6%.
Key Findings
- Best rate rose from 2.5% at launch (~Oct 2025) to 16.1% in under eight months — >4× improvement
- Prior leader: Opus 4.6 + Claude Cowork scaffold at 4.17%
- AI judges over-score 2.5–3× vs human evaluators — automated evaluation unreliable for hands-on software tasks
16% automation rate ≠ 16% job displacement. RLI measures task completion quality, not labor market substitution.