AudioBench - SEAME Singapore English WER: leaderboard
AudioBench SEAME Singapore English word error rate slice measuring automatic speech-recognition transcription quality, where lower scores indicate fewer errors.
Metric: WER (%) (lower is better). Source: github.com. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Whisper Large-v3 | 53.77 | #1271 |
| 2 | Qwen2-Audio-7B-Instruct | 54.87 | #880 |
| 3 | Phi-4 Multimodal Instruct | 85.29 | #906 |
| 4 | SALMONN-7B | 101.9 | #1293 |
| 5 | Qwen-Audio-Chat | 105.57 | #1244 |
| 6 | Gemini 1.5 Flash | 111 | #618 |
| 7 | SeaLLMs-Audio-7B | 171.07 | #1018 |
Interactive version: theaggregate.ai/benchmark?slug=audiobench-seame-singapore-english-wer · How It Works · Data refreshed daily, snapshot 2026-10-20.