MCGA - Speech Translation: leaderboard
Translating recited classical Chinese literature into English from audio (MCGA; Du et al., Harbin Institute of Technology, 2026), 1,000 test recordings. A DeepSeek-V3.2 judge rates the beauty of form, meaning and sound of each translation; the score is their mean on a 0-100 scale, higher is better.
Metric: LLM-B (0-100; mean of beauty of form, meaning and sound of the English translation of 1,000 test recordings, DeepSeek-V3.2 judge). Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 74 | #159 |
| 2 | Qwen3 Omni 30B A3B Instruct | 67.6 | #902 |
| 3 | Qwen2.5-Omni-7B | 49.7 | #1797 |
| 4 | Voxtral-Small-24B-2507 | 34.1 | #1291 |
| 5 | Phi-4 Multimodal Instruct | 27.5 | #3914 |
| 6 | Voxtral-Mini-3B-2507 | 25.5 | #3702 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mcga-speech-translation · How It Works · Data refreshed daily, snapshot 2026-11-12.