CarryOnBench - Harmfulness (Turn 1): leaderboard
CarryOnBench (2026) starts from seemingly harmful requests that have a benign purpose and checks whether a model gives the benign information a user needs, and how harmful its answers get, as simulated users clarify or push back. Judge-rated harmfulness from 1 to 5 in the first response to the bare request, taking the worst turn; lower is better.
Metric: Harmfulness (1-5): GPT-5.2 rates each response for harmful or policy-violating content (4 or more is harmful); a conversation scores its worst turn, averaged over conversations, over the first response to the bare query, on CarryOnBench (398 seemingly harmful SORRY-Bench queries with a benign underlying intent; user turns simulated by Gemini-3-Flash); lower is better. Source: arxiv.org. Saturation forecast: Not forecast. 14 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | OLMo 3.1 32B Instruct | 1.3 | #2509 |
| 2 | GPT-5.4 | 1.4 | #138 |
| 3 | GPT-OSS-120B | 1.4 | #712 |
| 4 | Claude Haiku 4.5 | 1.5 | #587 |
| 5 | GPT-5 Mini | 1.5 | #373 |
| 6 | Claude Sonnet 4 | 1.5 | #414 |
| 7 | Qwen 3 235B A22B | 1.5 | #633 |
| 8 | Qwen 3 Next 80B A3B Instruct | 1.5 | #781 |
| 9 | OLMo 3 7B Instruct | 1.5 | #6953 |
| 10 | Llama 4 Maverick | 1.8 | #1113 |
| 11 | Gemini 3.1 Pro (Preview) | 2 | #102 |
| 12 | DeepSeek V3.1 | 2.2 | #511 |
| 13 | Qwen 3 32B | 2.3 | #983 |
| 14 | Mixtral 8x7B Instruct | 2.8 | #4692 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=carryonbench-harmfulness-turn-1 · How It Works · Data refreshed daily, snapshot 2026-11-22.