KOpenAudioBench - KLlamaQ: leaderboard
Spoken Korean short-answer questions transferred from LLaMA Questions, part of KOpenAudioBench (Kim et al., 2026). Accuracy in percent over 284 TTS-synthesized items; higher is better.
Metric: Accuracy (%; short-answer questions transferred from LLaMA Questions; Korean spoken questions synthesized with TTS; 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 | Raon-Speech | 62.7 | |
| 2 | HyperCLOVA X 8B Omni | 60.2 | |
| 3 | Fun-Audio-Chat | 51.8 | |
| 4 | Qwen2.5-Omni | 47.9 | #1499 |
| 5 | MiniCPM-o 4.5 | 38.4 | |
| 6 | Step-Audio-2-mini | 35.9 | |
| 7 | Audio Flamingo 3 | 11.3 | |
| 8 | InteractiveOmni | 9.5 |
Interactive version: theaggregate.ai/benchmark?slug=kopenaudiobench-kllamaq · How It Works · Data refreshed daily, snapshot 2026-11-12.