Hire Fairly Resume Audit: leaderboard
The model decides callbacks for 24,024 pairs of identical resumes per model that differ only in a Black- or White-coded name; the score is the absolute gap in callback rate in percentage points. Unlike FairLens it measures one decision with paired counterfactual resumes, and either direction of preference counts.
Metric: Absolute race selection gap between paired resumes (percentage points). Source: arxiv.org. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 Mini | 0.14 | #389 |
| 2 | Llama 3.3 70B Instruct | 0.16 | #1357 |
| 3 | GPT-OSS-120B | 0.16 | #700 |
| 4 | Claude 3 Haiku | 0.35 | #2609 |
| 5 | Gemini 2.5 Flash | 0.37 | #453 |
| 6 | Grok 4.1 Fast | 0.4 | #430 |
| 7 | GPT-4o Mini | 0.61 | #1623 |
| 8 | Gemma 4 31B (IT) | 0.67 | #342 |
| 9 | Claude Haiku 4.5 | 0.7 | #568 |
| 10 | Qwen 3.5 397B A17B | 1.04 | #246 |
| 11 | DeepSeek V3.1 | 1.04 | #497 |
| 12 | GLM-5.1 | 1.6 | #179 |
| 13 | GPT-3.5 Turbo | 2.12 | #2753 |
| 14 | Llama 3.1 8B Instruct | 3.01 | #4246 |
Interactive version: theaggregate.ai/benchmark?slug=hire-fairly-resume-audit · How It Works · Data refreshed daily, snapshot 2026-11-13.