KIDBench (No Cue) - Boundary-setting: leaderboard
Metric: Boundary-setting score (1-5): whether the response sets age-appropriate limits around unsafe, private or adult topics and points the child to a safe adult when needed, rated 1-5 by a DeepSeek-V4-Pro judge with the KIDBench rubric and category-specific rules, temperature 0 for responder and judge, averaged on the KIDBench child-safety prompts for children aged 7 to 11 (50 per category, 10 categories), single-turn, no-cue prompts (no sign that the user is a child); higher is better. Source: arxiv.org. Saturation forecast: Around 2033. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Flash | 2.97 |
| 2 | Qwen 3.6 27B | 2.87 |
| 3 | Llama 3.3 70B Instruct | 2.82 |
| 4 | Gemma 3 12B | 2.8 |
| 5 | GPT-5 Mini | 2.79 |
| 6 | Qwen 3.5 4B | 2.77 |
| 7 | Gemini 3.1 Flash Lite | 2.71 |
| 8 | Claude Haiku 4.5 | 2.69 |
| 9 | Gemma 4 31B (IT) | 2.66 |
| 10 | Qwen 3 8B | 2.6 |
| 11 | Gemma 3 4B | 2.52 |
| 12 | Llama 3.1 8B Instruct | 2.45 |
Interactive version: theaggregate.ai/benchmark?slug=kidbench-no-cue-boundary-setting · How It Works · Data refreshed daily, snapshot 2026-10-07.