Best Korean Language Models
Comparing South Korea's top sovereign AI models on domestic benchmarks including KMMLU, CLIcK, and HRM8K. These models are optimized for Korean language tasks and cultural context.
Solar currently leads this regional leaderboard with an average Korean score of 80.1.
This page is intentionally regional-only. If you want to compare these models against GPT, Claude, Gemini, and other global systems, use the cross-market Korean benchmarks leaderboard.
| Rank | Model | KMMLU | Avg. Korean Score |
|---|---|---|---|
| #1 | Solar 128K•Proprietary | 80.1% | 80.1 |
| #2 | HyperClova X 128K•Open Weight | 78.4% | 78.4 |
| #3 | K-Exaone 256K•Proprietary | — | 76.0 |
| #4 | Exaone 4.0 128K•Open Weight | 75.2% | 75.2 |
What this page ranks
Only Korean-market or Korea-specific models are included here. The score is a straight average of available Korean benchmark rows, not BenchLM's global weighted score.
What to use next
If your decision is really “regional specialist or global frontier model,” move to the mixed benchmark view and compare both on the same Korean tasks.
Open Korean BenchmarksKorean leaderboard updates
Get notified when regional models move on KMMLU, CLIcK, and related Korean benchmarks.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.