Benchmark checker Law, finance, health, science

Is LegalBench still meaningful?

Saturated Checked against the report, 5 November 2026

No. The top ten models sit within 1.6 points, and it covers US law only.

Why saturated: The chapter rates it Saturated: the top ten models span only 1.6 points.^{[1][1] LegalBench Vals AI Primary}

Known problems

It covers US law only; the repository saw its last push in March 2026.^{[3][3] LegalBench repository GitHub Primary}
It tests legal skill, not lawful conduct.^{[4][4] Benchmark list Vals AI Primary}

HLAB-Vals 1.1 Status:Still useful Yes. All-pass scoring is strict, so models still spread out.
Harvey Legal Agent Bench (LAB) Status:Contested Only with the host named. The same model scores more than twice as high on another board.

Chapter 1, Benchmarks

Copied from the chapter source lists. Each number links back to the source in its chapter.

[1]
LegalBench, Vals AI. It gives current LegalBench scores by model. Primary vals.ai Chapter 1, source 66
[2]
LegalBench, arXiv 2308.11462. It describes 162 legal reasoning tasks. Primary arxiv.org Chapter 1, source 67
[3]
LegalBench repository, GitHub. It shows the date of the last push. Primary github.com Chapter 1, source 68
[4]
Benchmark list, Vals AI. It lists Vals's legal, finance, health, science and cyber boards. Primary vals.ai Chapter 1, source 15