KMMAU - General Counting: leaderboard
Counting questions about the spoken content, a four-way choice, on native Korean speech in KMMAU (Kim et al., 2026). Accuracy in percent; higher is better.
Metric: Accuracy (%; four-choice counting question; native Korean audio from KSS, KMSAV and the Seoul Corpus; single run at temperature 0.7). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Fun-Audio-Chat | 69.2 | |
| 2 | Raon-Speech | 65.4 | |
| 3 | MiniCPM-o 4.5 | 42.3 | |
| 4 | Step-Audio-2-mini | 40.4 | |
| 5 | Qwen2.5-Omni | 38.5 | #1499 |
| 6 | Audio Flamingo 3 | 34.6 | |
| 7 | HyperCLOVA X 8B Omni | 28.8 | |
| 8 | InteractiveOmni | 25 |
Interactive version: theaggregate.ai/benchmark?slug=kmmau-general-counting · How It Works · Data refreshed daily, snapshot 2026-11-12.