Legal BenchmarksWhat this report coversA summary of Acme Legal AI’s results across contract drafting and data extraction, including strengths, areas for improvement, and five certification categories.
Methodology & full resultsSee how tasks, scoring, comparison groups, and repeat runs work on the methodology page, then explore every measure in the Full Evaluation Report.
Provisional Q3 results. Q3 scores and ratings are provisional and may change during validation. Final results will be confirmed on 15 August 2026.
Report date:
Creating, amending, redlining, transforming, or completing contractual language or instruments.
Extracting, comparing, or synthesising facts, clauses, obligations, dates, entities, or terms from documents.
Lawyer overview of navigation, files, responsiveness, outputs, and continuity.
Biggest differentiator compared to other Legal AI tools
Extracting, comparing, or synthesising facts, clauses, obligations, dates, entities, or terms from documents. Acme Legal AI ranked first across every system evaluated, including frontier general-purpose AI.
Where Acme Legal AI performs most strongly against other legal AI
Acme Legal AI completed the same task successfully in both independent runs more often than any other system evaluated, legal or general-purpose.
Acme Legal AI got governing rules, obligations, and risk positions right more often than any other legal AI application.
Lawyer overview of navigation, files, responsiveness, outputs, and continuity.
Where Acme Legal AI stands weakest against other legal AI, and what reaching the level its peers already hold would be worth.
Whether the requested file type and output format were delivered.
An execution-ready final arrived with formatting residue: bracketed placeholder text, a stray punctuation mark, and unresolved blanks in signature fields, instead of a clean final with every field resolved. On other tasks a requested two-file package arrived as one combined document.
Whether answers stay grounded in the supplied facts and documents.
Asked for the consideration payable under a phased purchase agreement, Acme Legal AI reported one grand total the document never states: it summed the per-phase amounts and presented the sum as a term of the agreement, instead of reporting each component as stated and labeling any computed figure as derived.
Performance on scanned, photographed, image-based, redacted, or otherwise imperfect source material.
A scanned letter of intent stated per-phase acreage figures with poor print quality. Acme Legal AI reported a materially different set of acreages as settled figures, instead of flagging that the readings were unclear.
Independent certifications for performance measured against published legal AI application benchmarks.
Contender · Q3 2026
Contender · Q3 2026
Band 1 · Q2 2026
Top 30% of this vertical
Contender · Q3 2026
Contender · Q3 2026