CarryOnBench - Utility (No Clarifications): leaderboard
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.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 Next 80B A3B Instruct | 49.6 |
| 2 | GPT-5 Mini | 48.6 |
| 3 | Qwen 3 32B | 48.2 |
| 4 | DeepSeek V3.1 | 44.1 |
| 5 | GPT-5.4 | 42.8 |
| 6 | Qwen 3 235B A22B | 40.2 |
| 7 | OLMo 3 7B Instruct | 34.2 |
| 8 | Olmo 3.1 32B Instruct | 33.7 |
| 9 | Gemini 3.1 Pro (Preview) | 31.4 |
| 10 | Claude Sonnet 4 | 31.1 |
| 11 | Claude Haiku 4.5 | 28.5 |
| 12 | Mixtral 8x7B Instruct (v0.1) | 27.8 |
| 13 | Llama 4 Maverick | 17.8 |
| 14 | GPT-OSS-120B | 15.3 |
Interactive version: theaggregate.ai/benchmark?slug=carryonbench-utility-no-clarifications · How It Works · Data refreshed daily, snapshot 2026-10-07.