KIDBench (No Cue, Mandarin): leaderboard

KIDBench No Cue Mandarin asks the 500 KIDBench child-safety questions, translated into Mandarin and without any sign that the user is a child, and has the model answer in Mandarin (KIDBench, 2026). A DeepSeek-V4-Pro judge rates five child-safety dimensions from 1 to 5; the board shows their mean, higher is better.

Metric: Total 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, on the 500 no-cue KIDBench prompts translated into Mandarin by GPT-5.5 (system instruction sets the response language), averaged over the 10 categories, single-turn; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1DeepSeek V4 Flash3.29#267
2Claude Haiku 4.53.24#583
3Gemma 3 12B3.21#2059
4Qwen 3.5 4B3.19#1241
5Qwen 3.6 27B3.17#351
6Gemma 4 31B (IT)3.11#350
7Llama 3.3 70B Instruct3.07#1385
8GPT-5 Mini3.05#364
9Gemini 3.1 Flash Lite3.05#371
10Gemma 3 4B2.87#5339
11Qwen 3 8B2.86#1873
12Llama 3.1 8B Instruct2.58#4600
13KIDBench Llama-3.2-3B (checkpoint unspecified)2.25

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=kidbench-no-cue-mandarin · How It Works · Data refreshed daily, snapshot 2026-11-21.