CarryOnBench - Utility (Oracle Intent): 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 a single turn in which the query states the full benign intent, 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: Around 2030. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 Mini | 72.1 |
| 2 | Qwen 3 Next 80B A3B Instruct | 70.7 |
| 3 | DeepSeek V3.1 | 67.6 |
| 4 | GPT-5.4 | 63.2 |
| 5 | Qwen 3 32B | 61.2 |
| 6 | Qwen 3 235B A22B | 56.8 |
| 7 | Olmo 3.1 32B Instruct | 56.8 |
| 8 | Gemini 3.1 Pro (Preview) | 52.1 |
| 9 | Claude Sonnet 4 | 48.8 |
| 10 | OLMo 3 7B Instruct | 48.8 |
| 11 | Claude Haiku 4.5 | 45.9 |
| 12 | Mixtral 8x7B Instruct (v0.1) | 40 |
| 13 | Llama 4 Maverick | 35.7 |
| 14 | GPT-OSS-120B | 25.2 |
Interactive version: theaggregate.ai/benchmark?slug=carryonbench-utility-oracle-intent · How It Works · Data refreshed daily, snapshot 2026-10-07.