AudioBench
Audio LLM evaluation: ASR, speech translation, spoken QA, audio scene and paralinguistic tasks
speechactive
Definition
- Version
- not in the notes
- Category
- speech
- Maintainer
- A*STAR (Institute for Infocomm Research, I2R)
- Status
- active
- Home page
- https://huggingface.co/spaces/MERaLiON/AudioBench-Leaderboard
Sources
- huggingface.co/MERaLiON/MERaLiON-2-10B https://huggingface.co/MERaLiON/MERaLiON-2-10B
Scores at a Glance
Score in WER
Lower is better for this unit.
1MERaLiON-2-3BA*STAR (Institute for Infocomm Research, I2R)Indonesian ASR, average WER by languageMERaLiON-2-10BA*STAR (Institute for Infocomm Research, I2R)Indonesian ASR, average WER by language
MERaLiON-2-3BA*STAR (Institute for Infocomm Research, I2R)Indonesian ASR, average WER by language0.132 WER
vendor-reported2MERaLiON-2-10BA*STAR (Institute for Infocomm Research, I2R)Indonesian ASR, average WER by language0.143 WER
vendor-reported00.050.10.15
Where this comes from
Each bar is one model, with its most independent recorded score. Hatched bars are vendor-reported; the word beside each bar says who ran it. Settings and dates differ between rows; the Scores table below has every result.
Scores
| Model | Score | Who ran it | Setting | View or population | Date | Source |
|---|---|---|---|---|---|---|
| MERaLiON-2-10B | 0.143 WER | vendor-reported | Indonesian ASR, average WER by language | – | – | huggingface.co |
| MERaLiON-2-3B | 0.132 WER | vendor-reported | Indonesian ASR, average WER by language | – | – | huggingface.co |
Where this comes from
Scores from different settings, harnesses or dates are not directly comparable; the row says which. Sorted by score, highest first, as a reading aid, not a ranking.
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "AudioBench", https://aurion.semai.id/benchmarks/audiobench/, records of October 2026.