Enterprise AI in Numbers · Capability and cost
Top score on the Remote Labor Index: paid freelance projects AI completes acceptably
21.3%
21.3% (Claude Opus 5.5) as of September 22, 2026, according to Epoch AI, Benchmarking hub, up from 20.8% on September 3, 2026. The highest share of Remote Labor Index projects (real freelance jobs with their original client briefs) that an AI agent completes at an acceptable quality, with the model and lab.
What it means for leaders
The gap between benchmark scores and end-to-end jobs is still wide: most real projects aren't yet delivered acceptably without people. Plan agents into parts of workflows, with human review, rather than whole roles.
What would change the reading: Each new leader: the share rises in steps as frontier models ship.
The trend
Every reading
| Date | Value | Change | Detail | Source |
|---|---|---|---|---|
| 21.3% | +0.5 points | Claude Opus 5.5 | Source | |
| 20.8% | +2.9 points | GPT-6 Astra | Source | |
| 17.9% | +1.8 points | Claude Fable 5.1 | Source | |
| 16.1% | +7.8 points | Claude Fable 5 | Source | |
| 8.3% | +2.0 points | Claude Opus 4.8 | Source | |
| 6.3% | +2.1 points | GPT-5.5 | Source | |
| 4.2% | +0.4 points | Claude Opus 4.6 | Source | |
| 3.8% | +1.7 points | Claude Opus 4.5 | Source | |
| 2.1% | +0.4 points | Claude Sonnet 4.5 | Source | |
| 1.7% | +0.9 points | GPT-5 | Source | |
| 0.8% | — | Gemini 2.5 Pro (Jun 2025) | Source |
Definition and method
- What it measures
- The highest share of Remote Labor Index projects (real freelance jobs with their original client briefs) that an AI agent completes at an acceptable quality, with the model and lab.
- Population
- AI agents evaluated on the Remote Labor Index, as compiled by Epoch AI
- Updated
- When new models are evaluated
- Source
- Epoch AI, Benchmarking hub · CC BY 4.0
Caveats
- Scores come from independent evaluators compiled by Epoch AI, not from the labs, and depend on the settings and tools each run used.
- The leader is the best result so far, dated by the release of the model that set it; a model evaluated late can take the lead after its release date.
- Projects come from freelance marketplaces, so they over-represent self-contained digital work.
Cite this
Enterprise AI in Numbers, "Top score on the Remote Labor Index: paid freelance projects AI completes acceptably", https://lucianoayres.com/ai-brief/data/remote-work-benchmark-leader/, retrieved October 25, 2026. Data: Epoch AI, Benchmarking hub.