GenderBench - BBQ Gender Stereotype Rate: leaderboard

The model answers BBQ gender-identity questions; the number is the report's rate of answers that attribute a stereotyped behaviour to the associated gender. Unlike BBQ accuracy boards, it counts only stereotyped answers.

Metric: Rate of stereotypical answers on ambiguous BBQ gender items (0-1 normalized, lower is better). Source: genderbench.readthedocs.io. 12 models tracked.

Top models

#ModelScoreOverall rank
1GPT-4o0#741
2Gemma 2 9B (IT)0.01#2262
3Gemini 2.0 Flash0.01#704
4Phi-40.02#1913
5Gemma 2 27B (IT)0.02#1782
6Gemini 2.0 Flash Lite0.03#1032
7Llama 3.3 70B Instruct0.04#1357
8Mistral Small 30.05#1777
9GPT-4o Mini0.08#1623
10Claude 3.5 Haiku0.1#1553
11Llama 3.1 8B Instruct0.21#4246
12Mistral 7B Instruct (v0.3)0.24#6653

Interactive version: theaggregate.ai/benchmark?slug=genderbench-bbq-gender-stereotype-rate · How It Works · Data refreshed daily, snapshot 2026-11-13.