SEA-SpeechBench - Speech Translation (English Prompt): leaderboard
Speech translation into English in SEA-SpeechBench (Liao et al., 2026), from short Southeast Asian speech clips, with an English instruction prompt. Corpus BLEU from 0 to 100; higher is better.
Metric: BLEU (0-100): corpus BLEU of the English translation of Southeast Asian speech; short clips of at most 30 s from curated public Southeast Asian speech corpora, up to 1,000 sampled items per dataset; English instruction prompt. Source: arxiv.org. Saturation forecast: Around 2029. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Voxtral-Mini-3B-2507 | 19.98 | #3702 |
| 2 | MERaLiON-2-10B | 17.75 | #1646 |
| 3 | Gemini 2.5 Flash | 16.86 | #452 |
| 4 | Qwen3 Omni 30B A3B Instruct | 15.28 | #902 |
| 5 | Gemma 3n E4B (IT) | 10.98 | #2220 |
| 6 | Gemma 3n E2B (IT) | 8.97 | #3758 |
| 7 | Qwen2.5-Omni-7B | 7.91 | #1797 |
| 8 | MERaLiON-2-3B | 7.56 | #6258 |
| 9 | Phi-4 Multimodal Instruct | 3.04 | #3914 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sea-speechbench-speech-translation-english-prompt · How It Works · Data refreshed daily, snapshot 2026-11-12.