LongAudioSpan - Accuracy: leaderboard
LongAudioSpan multiple-choice accuracy on English and Chinese recordings of 10 minutes to over 2 hours, averaged over the short, medium and long duration tiers. Questions are drawn from the audio's own content and span perception, understanding and reasoning.
Metric: Accuracy (%, mean over S/M/L tiers). Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 Omni Plus | 74.6 | #111 |
| 2 | Gemini 3.1 Pro (Preview) | 72.3 | #64 |
| 3 | Qwen3.5 Omni Flash | 68.1 | #225 |
| 4 | Gemini 3 Flash | 66.3 | #103 |
| 5 | Gemini 2.5 Pro | 62.1 | #159 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=longaudiospan-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-05.