MCGA - Spoken QA: leaderboard
Open-ended questions about recited classical Chinese literature, answered from the audio (MCGA; Du et al., Harbin Institute of Technology, 2026), 1,000 test recordings. F1 against the reference answers in percent; higher is better.
Metric: F1 (%; open-ended answers to questions about 1,000 test recordings). Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3 Omni 30B A3B Instruct | 51.5 | #902 |
| 2 | Gemini 3 Flash | 48.7 | #159 |
| 3 | Qwen2.5-Omni-7B | 43.5 | #1797 |
| 4 | Voxtral-Small-24B-2507 | 27.9 | #1291 |
| 5 | Phi-4 Multimodal Instruct | 24.5 | #3914 |
| 6 | Voxtral-Mini-3B-2507 | 12.5 | #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-spoken-qa · How It Works · Data refreshed daily, snapshot 2026-11-12.