Legal Hallucination Detection Benchmark (no ex: leaderboard
Attorneys, judges, and pro se filers increasingly use AI to draft legal documents, yet these tools frequently fabricate citations. Despite predictions that newer models would hallucinate less or that court sanctions would deter negligent filers, we found over 1,000 filings containing fabricated citations -- with this number growing year-over-year. This study evaluates whether AI-based systems can
Metric: F1 (%) - Agentic (self-reported). Source: benchmarklist.com. Saturation forecast: Around July 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.8 | 68.8 | #37 |
| 2 | GPT-5 | 60.5 | #98 |
| 3 | Qwen 3.5 27B | 38 | #187 |
| 4 | GPT-OSS-120B | 34.4 | #341 |
| 5 | Gemini 2.5 Flash | 30.9 | #245 |
| 6 | Qwen 3 8B | 22.9 | #674 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=legal-hallucination-detection-benchmark-no-ex · How It Works · Data refreshed daily, snapshot 2026-10-20.