KoALa-Bench - CLIcK: leaderboard

KoALa-Bench (2026) evaluates large audio language models on Korean speech understanding and on faithfulness to the speech context, under clean and noisy audio. This board is accuracy in percent on spoken multiple-choice questions about Korean culture and language from CLIcK, clean audio; higher is better.

Metric: Accuracy (%) on the spoken short-form CLIcK multiple-choice questions on Korean culture and language, Korean speech synthesized or recorded per item, clean audio (first value of each clean / noisy pair), mean over four prompts with greedy decoding; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Flash Lite67.27#964
2Qwen3 Omni 30B A3B Instruct63.57#930
3GPT-Audio Mini61.64#918
4Raon-Speech-9B55.68
5Voxtral-Mini-3B-250741.94#3682
6Gemma 3n E4B (IT)35.11#2401

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=koala-bench-click · How It Works · Data refreshed daily, snapshot 2026-11-21.