GlobeAudio: leaderboard
Metric: Macro-average over six languages of accuracy (%) on GlobeAudio's naturalistic-audio multiple-choice questions written by native speakers; audio fed to the model directly, source-language prompts, predicted option label against the key. Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 82.68 |
| 2 | Gemini 2.0 Flash | 66.72 |
| 3 | MiMo-V2.5 | 61.08 |
| 4 | GPT-4o Audio | 60.13 |
| 5 | Voxtral-Small-24B-2507 | 57.94 |
| 6 | Gemma 3n E4B (IT) | 56.53 |
| 7 | MERaLiON-2-10B | 55.64 |
| 8 | Qwen2-Audio-7B-Instruct | 42.31 |
Interactive version: theaggregate.ai/benchmark?slug=globeaudio · How It Works · Data refreshed daily, snapshot 2026-09-29.