GPT-4o-mini-Audio (Preview 2024-12-17): benchmark results
Provider: OpenAI. Access: API.
Aggregate Rating 1496 ± 38, rank #2418 of 10916 rated models, from 34 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM Audio - Mutox | 87.39 | EM | 100 |
| HELM Audio - Toxicity Detection | 100 | Mean win rate | 100 |
| HELM Audio - Bias | 87.5 | Mean win rate | 87.5 |
| HELM Audio - Mustard | 62.32 | EM | 87.5 |
| HELM Audio - Parade | 85.67 | EM | 87.5 |
| HELM Audio - Safety | 68.75 | Mean win rate | 68.8 |
| HELM Audio - Vocal Sound | 79.4 | PEM | 68.8 |
| HELM Audio - Voice Jailbreak Attacks | 96.67 | Refusal rate for safety | 68.8 |
| HELM Audio - Audiocaps | 1.83 | GPT-4o Judge Critique | 50 |
| HELM Audio - Fleurs | 54.48 | WER | 50 |
| HELM Audio - Air Bench Foundation | 54.1 | EM | 43.8 |
| HELM Audio - Fleurs Fairness | 5.2 | WER | 43.8 |
Interactive version: theaggregate.ai/model?slug=gpt-4o-mini-audio-preview-2024-12-17 · How It Works · Data refreshed daily, snapshot 2026-11-08.