AudioBench - IMDA 30s SQA Human: leaderboard
AudioBench IMDA 30s SQA Human slice measuring audio-language understanding for the named dataset or task family.
Metric: Mean Score (%). Source: github.com. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | cascade_whisper_large_v2_gemma2_9b_cpt_sea_lionv3_instruct | 62.95 | |
| 2 | cascade_whisper_large_v3_llama_3_8b_instruct | 58.62 | |
| 3 | Qwen2-Audio-7B-Instruct | 47.1 | #880 |
| 4 | SALMONN_7B | 42.3 | |
| 5 | Qwen-Audio-Chat | 42.2 | #1244 |
Interactive version: theaggregate.ai/benchmark?slug=audiobench-imda-30s-sqa-human · How It Works · Data refreshed daily, snapshot 2026-10-20.