Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Agentic benchmark
Legal Agent Benchmark mean criterion-pass rate — Harvey held-out set. Every model the catalog carries a published LAB criterion-pass (Harvey held-out) value for, ranked by that value.
Harvey AI's mean criterion-level score on its held-out legal-agent evaluation. 3 models in the ModelScale catalog have a published value.
Each point is a model with a published LAB criterion-pass (Harvey held-out) value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
3 models with a published LAB criterion-pass (Harvey held-out) value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Mean criterion-pass rate |
|---|---|---|---|
| 1 | Anthropic | 94.1 | |
| 2 | Anthropic | 93.1 | |
| 3 | Anthropic |
What the benchmark measures and how its value is published, from the source’s own record.
LAB criterion-pass (Harvey held-out) (Legal Agent Benchmark mean criterion-pass rate — Harvey held-out set) is an agentic benchmark published by Claude Opus 5 System Card. Harvey AI's mean criterion-level score on its held-out legal-agent evaluation. Its evaluation set is described by the source as harvey-held-out legal-agent tasks. The source rates its difficulty as professional legal work.
Each model's result is published by Claude Opus 5 System Card as mean criterion-pass rate. ModelScale prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 3 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
Evidence key: Observed
Rows are ordered by the value Claude Opus 5 System Card published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.