CarryOnBench - Utility (No Clarifications): 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. Share of the benign information checklist fulfilled in multi-turn conversations where the user only pushes back, rephrases or makes non-contentful replies, in percent; higher is better.

Metric: Ben-Util (%): share of the benign information-need checklist (about 18 atomic yes/no items per query, built from five models' compliant answers) that the responses fulfil, each item judged separately by Gemini-2.5-Flash, over multi-turn conversations of 4 to 12 turns with only pushback, rephrasing and non-contentful user turns, conversation-level, on CarryOnBench (398 seemingly harmful SORRY-Bench queries with a benign underlying intent; user turns simulated by Gemini-3-Flash); higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 14 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 Next 80B A3B Instruct49.6#778
2GPT-5 Mini48.6#364
3Qwen 3 32B48.2#972
4DeepSeek V3.144.1#511
5GPT-5.442.8#135
6Qwen 3 235B A22B40.2#628
7OLMo 3 7B Instruct34.2#6953
8OLMo 3.1 32B Instruct33.7#2445
9Gemini 3.1 Pro (Preview)31.4#99
10Claude Sonnet 431.1#410
11Claude Haiku 4.528.5#583
12Mixtral 8x7B Instruct27.8#4696
13Llama 4 Maverick17.8#1110
14GPT-OSS-120B15.3#704

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=carryonbench-utility-no-clarifications · How It Works · Data refreshed daily, snapshot 2026-11-21.