Korean LLM Eval - KMMLU — leaderboard
Metric: Accuracy (%, 0-shot). Source: github.com. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.1 (Medium) | 83.73 |
| 2 | GPT-5 Mini | 76.47 |
| 3 | GPT-5.2 | 71.54 |
| 4 | GPT-5 Nano | 69.28 |
| 5 | GPT-5 Chat | 65.92 |
| 6 | GPT-5.1 | 65.9 |
| 7 | GPT-4.1 | 65.49 |
| 8 | GPT-4o | 64.26 |
| 9 | GPT-5.1 ChatGPT | 63.11 |
| 10 | GPT-4.1 Mini | 59.26 |
| 11 | GPT-4 Turbo | 58.75 |
| 12 | GPT-4o Mini | 52.63 |
| 13 | Phi-4 | 50.3 |
| 14 | GPT-4.1 Nano | 48.57 |
| 15 | Phi-3.5-MoE-instruct | 47.43 |
Interactive version: theaggregate.ai/benchmark?slug=korean-llm-eval-kmmlu · How It Works · Data refreshed daily, snapshot 2026-07-27.