AudioBench - SEAME Mandarin WER: leaderboard
AudioBench SEAME Mandarin word error rate slice measuring automatic speech-recognition transcription quality, where lower scores indicate fewer errors.
Metric: WER (%) (lower is better). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen2-Audio-7B-Instruct | 55.23 | #880 |
| 2 | Phi-4 Multimodal Instruct | 71.26 | #906 |
| 3 | Whisper Large-v3 | 72.26 | #1271 |
| 4 | Qwen-Audio-Chat | 87.83 | #1244 |
| 5 | Gemini 1.5 Flash | 96.91 | #618 |
| 6 | SeaLLMs-Audio-7B | 106.39 | #1018 |
| 7 | SALMONN-7B | 127.22 | #1293 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=audiobench-seame-mandarin-wer · How It Works · Data refreshed daily, snapshot 2026-10-20.