SALMONN-7B: benchmark results
Provider: Tsinghua & ByteDance. Released 2023-11-13. Access: Open.
Aggregate Rating 1380 ± 15, rank #1293 of 1656 rated models, from 96 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AudioBench - CoVoST2 TA-EN BLEU | 36.49 | Mean Score (%) | 100 |
| AudioBench - WavCaps QA | 47.3 | Mean Score (%) | 100 |
| AudioBench - AudioCaps QA | 50.29 | Mean Score (%) | 87.5 |
| BASS - Artist Counting | 27.32 | Exact-match accuracy (%) on the 229 artist counting question | 84.6 |
| AudioBench - Clotho AQA | 57.75 | Mean Score (%) | 75 |
| AudioBench - IEMOCAP Gender | 81.31 | Mean Score (%) | 75 |
| AudioBench - IMDA GR Sentence | 59.77 | Mean Score (%) | 75 |
| AudioBench - IMDA GR Dialogue | 42.73 | Mean Score (%) | 62.5 |
| AudioBench - VoxCeleb Accent | 34.22 | Mean Score (%) | 62.5 |
| AudioBench - VoxCeleb Gender | 88.8 | Mean Score (%) | 57.1 |
| AudioBench - GigaSpeech WER | 10.77 | WER (%) (lower is better) | 55.6 |
| AudioBench - IMDA Part 4 ASR WER | 77.57 | WER (%) (lower is better) | 55.6 |
Interactive version: theaggregate.ai/model?slug=salmonn-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.