FoodGuardBench - Vanilla Harmful Queries: leaderboard
Refusal of plainly worded harmful food-safety requests from FoodGuardBench (2026): 2,339 queries grounded in FDA food-code principles across eight hazard categories, asked without any jailbreak. The score is the attack success rate, the percent of requests answered without a refusal phrase; lower is better.
Metric: Attack success rate (%) of the plain, non-jailbroken harmful food-safety queries of FoodGuardBench (2,339 vanilla queries grounded in FDA food-code principles across eight hazard categories), temperature 0.7, a response counting as a successful attack when it contains none of a curated set of refusal prefixes; lower is better. Source: arxiv.org. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GLM-4 32B | 10 | #1325 | |
| 2 | Claude 3.7 Sonnet | 10.56 | #470 | |
| 3 | GPT-4.1 | 10.85 | #505 | |
| 4 | Qwen 3 8B | Non-reasoning | 13.16 | #1873 |
| 5 | Qwen 3 32B | Non-reasoning | 15.9 | #972 |
| 6 | Qwen 2.5 7B | 18.42 | #2929 | |
| 7 | GPT-4o | 19.66 | #751 | |
| 8 | Mistral Small 4 | 31.81 | #1229 | |
| 9 | Llama 3.3 70B Instruct | 32.65 | #1385 |
Interactive version: theaggregate.ai/benchmark?slug=foodguardbench-vanilla-harmful-queries · How It Works · Data refreshed daily, snapshot 2026-11-21.