FDARxBench (Oracle Passages) - Multi-Hop: leaderboard
Metric: Answer accuracy (x 100) on multi-hop questions from FDARxBench's expert-guided QA items grounded in 700 FDA prescription drug labels, where only the gold evidence passages are given as context, graded against the reference answer by a GPT-5.1 judge (correct only if all clinically important content is present and nothing contradicts it); higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o Mini | 82.2 | #588 |
| 2 | GPT-5.2 | 80.6 | #105 |
| 3 | GPT-5.1 | 80.5 | #131 |
| 4 | Qwen 3 32B | 79.6 | #424 |
| 5 | Qwen 3 14B | 78.9 | #524 |
| 6 | Claude Opus 4.6 | 72.8 | #60 |
| 7 | Llama 3.3 70B Instruct | 68 | #520 |
| 8 | Ministral-3-14B-Instruct-2512 | 67.6 | #590 |
| 9 | Llama 3.1 8B Instruct | 66.8 | #1018 |
| 10 | Claude Sonnet 4.5 | 65.3 | #138 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=fdarxbench-oracle-passages-multi-hop · How It Works · Data refreshed daily, snapshot 2026-10-11.