PolySpeech-100 - Chinese Dialects: leaderboard
Metric: Accuracy (%; spoken four-option reading comprehension from Belebele: the audio carries the passage, question and options; text instruction as system prompt; mean over 19 Chinese regional dialects). Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 83.54 |
| 2 | Qwen2.5-Omni-7B | 78.61 |
| 3 | MiniCPM-o-4.5 | 59.24 |
Interactive version: theaggregate.ai/benchmark?slug=polyspeech-100-chinese-dialects · How It Works · Data refreshed daily, snapshot 2026-09-26.