AudioBench - CoVoST2 ZH-EN BLEU: leaderboard
AudioBench CoVoST2 ZH-EN BLEU speech-translation BLEU slice measuring translation quality between the named source and target languages.
Metric: Mean Score (%). Source: github.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Phi-4 Multimodal Instruct | 22.68 | #906 |
| 2 | SeaLLMs-Audio-7B | 18.79 | #1018 |
| 3 | Qwen2-Audio-7B-Instruct | 16.47 | #880 |
| 4 | Whisper Large-v3 | 14.67 | #1271 |
| 5 | Qwen-Audio-Chat | 9.9 | #1244 |
| 6 | SALMONN-7B | 5.3 | #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-covost2-zh-en-bleu · How It Works · Data refreshed daily, snapshot 2026-10-20.