Legal Agent Benchmark mean criterion-pass rate — Harvey held-out set (LAB criterion-pass (Harvey held-out))
We show this table for reference; we do not rank on it.
Harvey AI's mean criterion-level score on its held-out legal-agent evaluation.
Benchmark score on LAB criterion-pass (Harvey held-out) — November 13, 2026
We compile the LAB criterion-pass (Harvey held-out) rows from provider self-reports. Claude Opus 5 leads the table at 94.1%, followed by Claude Sonnet 5.5 (93.1%) and Claude Opus 5.5 (91.2%). We do not use these results to rank models overall.
Claude Opus 5
Anthropic
Reported by AnthropicClaude Sonnet 5.5
Anthropic
Reported by AnthropicClaude Opus 5.5
Anthropic
Reported by Anthropic3 modelsAgenticCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (3 models)
All 3 results on this page were reported by the labs that built the models.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Opus 5AnthropicReported by Anthropic | 94.1% | Reported by Anthropic | Not reported | Closed |
| 2 | Claude Sonnet 5.5AnthropicReported by Anthropic | 93.1% | Reported by Anthropic | Not reported | Closed |
| 3 | Claude Opus 5.5AnthropicReported by Anthropic | 91.2% | Reported by Anthropic | Not reported | Closed |
- Lab-reported
- Benchmark-run
- Independent
Among the reported LAB criterion-pass (Harvey held-out) rows, Claude Opus 5 is first at 94.1%. The third row is 2.9 points behind. The broader top-10 range is 2.9 points, so many of the published results sit in a relatively narrow band.
3 models have been evaluated on LAB criterion-pass (Harvey held-out). The benchmark falls in the Agentic category. LAB criterion-pass (Harvey held-out) is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About LAB criterion-pass (Harvey held-out)
Year
2026
Tasks
Harvey-held-out legal-agent tasks
Format
Mean criterion-pass rate
Difficulty
Professional legal work
Section 8.13.3 separately reports Harvey's held-out criterion-pass result. It is not merged with Anthropic's internal-harness result.
Freshness and provenance
Version
LAB criterion-pass (Harvey held-out) 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does LAB criterion-pass (Harvey held-out) measure?
Harvey AI's mean criterion-level score on its held-out legal-agent evaluation.
Which model scores highest on LAB criterion-pass (Harvey held-out)?
Claude Opus 5 by Anthropic currently leads with a score of 94.1% on LAB criterion-pass (Harvey held-out).
How many models are evaluated on LAB criterion-pass (Harvey held-out)?
3 AI models have published results on LAB criterion-pass (Harvey held-out) in the BenchLM catalog.
Compare top models on LAB criterion-pass (Harvey held-out)
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.