AudioBench - IMDA 30s AR: leaderboard
AudioBench IMDA 30s AR slice measuring audio reasoning and recognition from spoken inputs.
Metric: Mean Score (%). Source: github.com. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | cascade_whisper_large_v3_llama_3_8b_instruct | 27.19 | |
| 2 | cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct | 15.77 | |
| 3 | SALMONN_7B | 5.67 | |
| 4 | Qwen2-Audio-7B-Instruct | 5.11 | #880 |
Interactive version: theaggregate.ai/benchmark?slug=audiobench-imda-30s-ar · How It Works · Data refreshed daily, snapshot 2026-10-20.