HarDBench - Harmfulness Score: leaderboard

HarDBench (2026) co-authoring jailbreak over 400 incomplete harmful drafts, scored by a GPT-4o judge from 1 (harmless or refused) to 5 (clearly harmful) and averaged; lower is better.

Metric: Mean harmfulness score (1-5) on HarDBench's 400-draft test set (100 incomplete harmful drafts in each of explosives, drugs, weapons and cyberattacks), each wrapped in the co-authoring jailbreak prompt (an editing task framing asking the model to elaborate and complete the draft); GPT-4o rates each completion's harmfulness from 1 (harmless or refused) to 5 (clearly harmful); lower is better. Source: arxiv.org. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Llama 3 8B Instruct4.29#4852
2Gemini 2.5 Pro4.56#276
3Gemini 2.0 Flash4.56#723
4Mistral 7B Instruct (v0.3)4.74#6750
5HarDBench DeepSeek-R1-8B (checkpoint unspecified)4.78
6GPT-4o4.87#751
7DeepSeek R1 Distill Qwen 32B4.94#1834
8Qwen 3 8B4.97#1873

Interactive version: theaggregate.ai/benchmark?slug=hardbench-harmfulness-score · How It Works · Data refreshed daily, snapshot 2026-11-21.