KIDBench (No Cue) - Benign Information Seeking: leaderboard
Metric: Category score (1-5): mean of the five child-safety dimensions, each rated 1-5 by a DeepSeek-V4-Pro judge with the KIDBench rubric and category-specific rules, temperature 0 for responder and judge, over the 50 Benign Information Seeking prompts of KIDBench for children aged 7 to 11, single-turn, no-cue prompts (no sign that the user is a child); higher is better. Source: arxiv.org. Saturation forecast: Around July 2028. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Flash | 4.07 |
| 2 | Llama 3.3 70B Instruct | 4.02 |
| 3 | Gemma 4 31B (IT) | 4.02 |
| 4 | Qwen 3.5 4B | 3.9 |
| 5 | Claude Haiku 4.5 | 3.84 |
| 6 | Qwen 3 8B | 3.84 |
| 7 | Gemini 3.1 Flash Lite | 3.84 |
| 8 | Qwen 3.6 27B | 3.82 |
| 9 | Gemma 3 4B | 3.74 |
| 10 | Llama 3.1 8B Instruct | 3.69 |
| 11 | Gemma 3 12B | 3.69 |
| 12 | GPT-5 Mini | 3.49 |
Interactive version: theaggregate.ai/benchmark?slug=kidbench-no-cue-benign-information-seeking · How It Works · Data refreshed daily, snapshot 2026-10-07.