GenderBench - GEST Creative Stereotype Rate: leaderboard
The model writes a character profile for a protagonist described by a GEST stereotype trait; the number is how much the protagonist's gender follows the stereotype, normalized as max(0, x). Unlike the GEST quote board, the model chooses the gender while writing.
Metric: Stereotyping of protagonists written for GEST traits, max(0, x) (0-1 normalized, lower is better). Source: genderbench.readthedocs.io. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.0 Flash | 0.11 | #704 |
| 2 | Claude 3.5 Haiku | 0.12 | #1553 |
| 3 | Gemma 2 9B (IT) | 0.13 | #2262 |
| 4 | Gemma 2 27B (IT) | 0.15 | #1782 |
| 5 | Gemini 2.0 Flash Lite | 0.18 | #1032 |
| 6 | Llama 3.3 70B Instruct | 0.2 | #1357 |
| 7 | Mistral Small 3 | 0.22 | #1777 |
| 8 | GPT-4o Mini | 0.23 | #1623 |
| 9 | Llama 3.1 8B Instruct | 0.23 | #4246 |
| 10 | Mistral 7B Instruct (v0.3) | 0.27 | #6653 |
| 11 | GPT-4o | 0.29 | #741 |
| 12 | Phi-4 | 0.34 | #1913 |
Interactive version: theaggregate.ai/benchmark?slug=genderbench-gest-creative-stereotype-rate · How It Works · Data refreshed daily, snapshot 2026-11-13.