RAIL - Auditory Processing: leaderboard
Metric: LLM-as-judge score (%) on auditory-processing tasks (speech-sound discrimination, distortion resistance, sound localization); GPT-5.4 judge; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 62.14 |
| 2 | Gemini 2.5 Flash | 53.76 |
| 3 | Gemini 3 Flash | 52.82 |
| 4 | GPT-4o Audio | 51.11 |
| 5 | MERaLiON-2-10B | 46.32 |
| 6 | Qwen2-Audio-7B-Instruct | 42.56 |
| 7 | Gemma 3n E4B (IT) | 36.75 |
| 8 | SALMONN-13B | 34.79 |
| 9 | GPT Audio | 26.32 |
Interactive version: theaggregate.ai/benchmark?slug=rail-auditory-processing · How It Works · Data refreshed daily, snapshot 2026-09-29.