Skip to content

Benchmark

Harvey Legal Agent Benchmark

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.

Models and systems measured
8
Scores recorded
8
Sources
1
IQ weight
No weight published

Score distribution

9%
Mean
10%
Median
6.1%
Std dev
2.5%
10th percentile
15.8%
90th percentile
0.8%
Min
0.8%19.6%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#Model or systemScore
1Grok 4.7Vals AI · 120-task held-out · Valkyrie · no internet · Harvey final score · xhigh effort· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.19.6%
2Gemini 3.8 Flashall-pass rate· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.10%
3Claude Sonnet 5Vals AI · 120-task held-out · Valkyrie · no internet · Harvey final score · max effort· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.5%
4GPT-5.6 SolVals AI · 120-task held-out · Valkyrie · no internet · Harvey final score · max effort· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.2.5%
5GPT-5.6 TerraVals AI · 120-task held-out · Valkyrie · no internet · Harvey final score · max effort· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.0.8%

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
Claude Fable 5 with Opus 4.8 fallbackVals AI · 120-task held-out · Valkyrie · no internet · Harvey final score · max effort· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.11.3%
Claude Fable 5.1 with Opus 5 fallbackVals AI · 120-task held-out · Valkyrie · no internet · Harvey final score · max effort· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.6.7%
Grok 4.6Vals AI · 120-task held-out · Valkyrie · no internet · Harvey final score · high effort· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.15.8%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.