Real legal work
Every task is a real legal problem that landed on a lawyer's desk, contributed by lawyers from more than 34 countries.
Application benchmarking
Get your application benchmarked against the industry standard
Independent, data-backed evidence of how your AI product performs in real legal work.
Certifications are issued in September 2026. Sign up by August 15 to be included.
Trusted by legal leaders across the world
How it works
- 01
Access.
Grant API access, or choose manual collection and evaluators work through your product interface.
- 02
Evaluation.
Your application completes the full task set under the same conditions as every benchmarked product.
- 03
Results and certification.
Your results are shared with your team. All benchmarked applications are eligible for quarterly certifications based on performance.
Benchmarked against real applications
Compared against legal AI applications and general-purpose applications like ChatGPT and Claude that lawyers already use for real legal work.
Assessed by lawyers
Outputs undergo automated grading and expert legal review for quality, judgment, and professional legal taste.
Assessed on
- Substance
- Form
- Product experience
Legal Benchmarks builds trust in legal AI by bringing testing closer to the real work of lawyers, with practical use cases and rigorous standards built for the profession.
Mohamed Al Mamari
Legal Counsel
Lawyers have the final say
Where automated grading and legal review disagree, the lawyer's verdict is binding and preserved for audit.
Read the methodologyLegal Benchmarks certification
Category Winner
Get certified
Every benchmarked product qualifies for the quarterly Legal Benchmarks certifications. Certifications are awarded on performance alone.
Next awards: September 2026
Pricing
Sign up by August 15, 2026 to be eligible for certifications issued this quarter (September 2026).
Public benchmark assessment
$500 per task category
API-based benchmark entry for commercial legal AI products.
- Named results on the public leaderboard
- Certification eligibility based on performance
- Automated grading and expert legal review
- No private assessment report
Private assessment
Custom pricing
A confidential, product-specific assessment with API-based output collection.
- Private product assessment report
- Exact strengths and failure modes
- Prioritized improvement areas
- Lawyer-rated output quality and usability
- Certification eligibility based on performance
The fee covers the evaluation itself. It has no effect on scores, leaderboard placement, or certification.
Built in public, with the people who buy legal AI.
This benchmark grew out of an open-source evaluation framework built with 100+ legal practitioners.
RESEARCH
Previous Application Benchmarks
Published application benchmark reports: data extraction and contract drafting studies of how AI performs on real legal work, openly available with full findings.
TRANSPARENCY
Leaderboard
Public rankings of general-purpose applications on the same tasks your product is tested on.
PROCESS
Methodology
Task design, grading, adjudication, retesting, and disclosure rules, documented end to end.
PEOPLE
Community
Tasks authored by practicing lawyers across 34 jurisdictions and growing, with an advisory board of GCs and legal ops leaders.
From anecdotal vibe to benchmark, Legal Benchmarks is valuable insight for the legal tech community.
Andrew Greenfeld
Legal Ops Leader
Frequently asked questions.
Fees, independence, and how results are handled.