LongAudioSpan - Rubric: leaderboard
LongAudioSpan rubric score on open-ended questions about long-form audio, graded by an LLM judge against per-question criteria that check both the content of an answer and the timestamps it cites, averaged over the three duration tiers.
Metric: Rubric score (%, mean over S/M/L tiers). Source: arxiv.org. Saturation forecast: Around May 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 54.6 | #103 |
| 2 | Qwen 3.5 Omni Plus | 50.8 | #111 |
| 3 | Gemini 3.1 Pro (Preview) | 49.4 | #64 |
| 4 | Qwen3.5 Omni Flash | 42.4 | #225 |
| 5 | Gemini 2.5 Pro | 41.3 | #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-rubric · How It Works · Data refreshed daily, snapshot 2026-11-05.