Grip on LLMs - HonestCityBench - Outdated Information: leaderboard
Grip on LLMs (arXiv 2608.09925v1, Appendix Table 2): HonestCityBench, 530 Dutch prompts that a model cannot or should not answer directly. Share of responses judged to appropriately acknowledge that the prompt needs information newer than the model has, percent. Higher is better.
Metric: Appropriate acknowledgement rate (%; LLM judge). Source: arxiv.org. 31 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Small 3 | 66 | #630 |
| 2 | Llama 3.1 8B Instruct | 62 | #1017 |
| 3 | GPT-4o | 61 | #357 |
| 4 | OLMo-2-0325-32B-Instruct | 53 | #847 |
| 5 | Phi-4 Mini Instruct | 52 | #1232 |
| 6 | EuroLLM-9B-Instruct | 43 | #1047 |
| 7 | GPT-4o Mini | 42 | #598 |
| 8 | GEITje-7B-ultra | 32 | #1435 |
| 9 | DeepSeek R1 Distill Qwen 32B | 27 | #662 |
| 10 | GPT-5 Mini | 15 | #193 |
| 11 | Qwen 3 32B | 14 | #449 |
| 12 | aya-expanse-32B | 14 | #687 |
| 13 | command-r7B-12-2024 | 14 | #1180 |
| 14 | Qwen 3 8B | 12 | #678 |
| 15 | GPT-5 Nano | 12 | #439 |
Interactive version: theaggregate.ai/benchmark?slug=grip-on-llms-honestcitybench-outdated-information · How It Works · Data refreshed daily, snapshot 2026-11-05.