HELM VHELM - Fairness: leaderboard
The "Fairness" subtask of HELM VHELM. Stanford HELM's VHELM project, the holistic evaluation of vision-language models: image-grounded question answering, OCR, chart and diagram reading, spatial reasoning, robustness, bias, fairness and toxicity scenarios scored under one ha...
Metric: Mean win rate. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.5 | 95.21 | #276 |
| 2 | GPT-4o (2024-05-13) | 90.43 | #326 |
| 3 | Gemini 2.0 Flash (Preview) | 88.83 | #366 |
| 4 | O3 (2025-04-16) | 82.98 | #123 |
| 5 | GPT-4.1 (2025-04-14) | 81.38 | #250 |
| 6 | GPT-4.1 Mini | 79.79 | #351 |
| 7 | Gemini 2.5 Pro (Preview 03-25) | 76.6 | #124 |
| 8 | GPT-4o (2024-08-06) | 72.34 | #335 |
| 9 | Gemini 2.0 Flash Lite | 72.34 | #453 |
| 10 | GPT-4o (2024-11-20) | 68.62 | #378 |
| 11 | Gemini 2.0 Flash | 68.09 | #344 |
| 12 | Gemini 1.5 Pro (002) | 68.09 | #350 |
| 13 | Gemini 1.5 Flash (002) | 64.36 | #585 |
| 14 | O4 Mini (2025-04-16) | 61.7 | #181 |
| 15 | Gemini 2.0 Pro (Preview 02-05) | 56.38 | #305 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-vhelm-fairness · How It Works · Data refreshed daily, snapshot 2026-10-23.