Counsel Stack offers multiple specialized LLMs for legal research, document review and client communications. In the VLAIR legal research study, their research product achieved the highest performance across all evaluation criteria, ranking first among all participants including legal AI products, generalist AI, and the lawyer baseline.

Performance Summary

Counsel Stack was evaluated on legal research capabilities across 200 U.S. legal research questions, with responses scored on three weighted criteria: accuracy (50% weight), authoritativeness (40% weight), and appropriateness (10% weight).

Counsel Stack demonstrated strong performance across all evaluated criteria, achieving the highest scores among all participants with 81% accuracy, 77% authoritativeness, and 71% appropriateness, resulting in a 78% aggregate weighted score.

Key Strengths

Based on VLAIR evaluation results, Counsel Stack demonstrates several notable capabilities:

1. Highest overall performance: Achieved the top score across all three evaluation criteria (accuracy, authoritativeness, and appropriateness) among all participants in the study, including other legal AI products, generalist AI, and the lawyer baseline.

2. Superior accuracy: At 81% accuracy, Counsel Stack achieved the highest accuracy score, demonstrating strong capability in providing substantively correct responses with minimal misinterpretations or factual errors.

3. Strong source citation: Scored 77% on authoritativeness, leading all participants in identifying and citing relevant and valid primary law sources, an essential differentiator for legal AI products over generalist alternatives.

4. Comprehensive legal coverage: Successfully provided responses to 196 of 200 questions, with only 4 technical timeout issues, demonstrating broad coverage of U.S. legal research questions across federal and state jurisdictions.

5. Specialized LLM architecture: Built with multiple specialized LLMs specifically designed for legal research, enabling the product to handle the unique requirements of legal source identification and citation.

6. Consistent quality: Outperformed the lawyer baseline by 9 percentage points overall (78% vs 69%), demonstrating the ability to augment and enhance legal research capabilities for practicing attorneys.

7. Superior to generalist AI on sources: Achieved significantly higher authoritativeness scores compared to generalist AI products, reflecting the value of access to proprietary legal databases and specialized training on legal sources.

Based on the Vals Legal AI Report (VLAIR) - Legal Research, October 2025 - Independent benchmarking study of legal AI products on U.S. legal research questions contributed by Am Law 100 law firms.