Remote Labor Index
Integrity rank #41 of 61 · 10 models scored · top score 16.1 · Claude Fable 5
unknown
Strongest on contamination resistance, weakest on discrimination. Caveat: the public test set does not rule out training-data contamination.
Reported scores — protocols may differ. How we compare
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | Claude Fable 5 | 16.1 | unverifiedT2 | 2026-06-09 |
| 2 | Claude Opus 4.8 | 8.33 | unverifiedT2 | 2026-05-28 |
| 3 | GPT-5.5 | 6.25 | unverifiedT2 | 2026-04-23 |
| 4 | Claude Opus 4.6 | 4.17 | unverifiedT2 | 2026-02-05 |
| 5 | Claude Opus 4.5 | 3.75 | unverifiedT2 | 2025-11-24 |
| 6 | GPT-5.2 | 2.5 | unverifiedT2 | 2025-12-11 |
| 7 | Claude Sonnet 4.5 | 2.08 | unverifiedT2 | 2025-09-29 |
| 8 | GPT-5 | 1.67 | unverifiedT2 | 2025-08-07 |
| 9 | Gemini 3 Pro | 1.25 | unverifiedT2 | 2025-11-18 |
| 10 | Gemini 2.5 Pro (Jun 2025) | 0.83 | unverifiedT2 | 2025-06-05 |
Continue from this leaderboard into the tightest head-to-head reads.
Weighted composite. Each dimension opens to the source-backed sub-signals it is computed from.
Claims drawn from cited facts, not live model generation.
This benchmark scores 10 models, with a spread of 15.27 points between the best and worst results. That spread is wide enough to suggest genuine separation between models, so rank differences likely reflect real capability gaps.
2 cited facts
The benchmark's Integrity Index is 81, graded B, ranking 41st out of 61 benchmarks. Its weakest integrity component is discrimination, which costs the reader by limiting how sharply the benchmark can separate frontier-model scores under its disclosed harness.
5 cited facts
The benchmark is not saturated, meaning the field at the top has not yet converged. The current top score leaves substantial headroom to the ceiling, and only a single model sits within a couple of points of the lead. Because it is not saturated, the ranking still offers genuine separation among the leading models, so small differences here reflect real capability gaps rather than noise.
5 cited facts
Because the disclosed harness is consistent, task performance scores are directly comparable across evaluations reported under the same harness. Test-set privacy is unknown, so whether this benchmark's test set is public or held out cannot be determined, leaving the appropriate scrutiny for possible contamination unspecified. Contamination history is unknown, and the harness notes only document reporting metadata, so no qualitative sensitivity or contamination risk can be inferred for this benchmark.
4 cited facts
This benchmark's design is not documented in the available record: what it measures, how it was constructed, its key assumptions, and its limitations are all marked unknown, and no human-baseline figure is provided. Because these core methodological facts are absent, the benchmark cannot tell a reader how to interpret its scores or where it is likely to mislead. The sharpest caveat to retain is that, based on the supplied facts, any claim about this benchmark's validity or comparison to human performance would be unsupported.
5 cited facts
Remote Labor Index: Remote Labor Index as reported in Epoch AI's Capabilities Index CSV.
Claude Fable 5 leads Remote Labor Index at 16.1. The full leaderboard above lists every recorded measurement, not just the headline number.
10 models have recorded scores on Remote Labor Index, spanning a score spread of 15.27.
tensor.news grades Remote Labor Index B for integrity (score 81/100), ranking #41 of 61 benchmarks we assess across discrimination, saturation, contamination resistance, harness comparability, and freshness.
No — Remote Labor Index still has headroom and continues to discriminate between models rather than bunching them at the ceiling.
Its test set is unknown, so contamination risk is on the table: a high score may partly reflect training-data overlap rather than capability.
Scores are reported under a consistent harness, so comparisons on Remote Labor Index are reasonably apples-to-apples.
Every Remote Labor Index measurement is source-backed and tagged reproduced, self-reported, or unverified; the underlying record cites epoch.ai/data/eci_benchmarks.csv.