StereoTales: leaderboard
Multilingual studies of social bias in open-ended LLM generation remain limited: most existing benchmarks are English-centric, template-based, or restricted to recognizing pre-specified stereotypes. We introduce StereoTales, a multilingual dataset and evaluation pipeline for systematically studying the emergence of social bias in open-ended LLM generation. The dataset covers 10 languages and 79 so
Metric: Emissions (self-reported). Source: benchmarklist.com. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Nano | 67 | #425 |
| 2 | Mistral Large 3 | 109 | #401 |
| 3 | Mistral Small 4 | 117 | #517 |
| 4 | Claude Haiku 4.5 | 121 | #280 |
| 5 | GPT-4o | 122 | #343 |
| 6 | GPT-4.1 | 133 | #248 |
| 7 | GPT-OSS-120B | 138 | #341 |
| 8 | MiniMax-M2.5 | 147 | #303 |
| 9 | Qwen 3.5 Flash | 157 | #232 |
| 10 | DeepSeek V3.2 | 158 | #205 |
| 11 | GPT-5.4 | 161 | #79 |
| 12 | GPT-5 Mini | 169 | #186 |
| 13 | Kimi K2.5 | 171 | #144 |
| 14 | Qwen 3.5 27B | 181 | #187 |
| 15 | Claude Sonnet 4.6 | 192 | #90 |
Interactive version: theaggregate.ai/benchmark?slug=stereotales · How It Works · Data refreshed daily, snapshot 2026-10-20.