Korean LLM Eval - KoBALT-700: leaderboard
daekeun-ml Korean proficiency suite on KoBALT-700: overall accuracy % on 700 items probing Korean linguistic competence across morphology, syntax and semantics, 0-shot.
Metric: Accuracy (%, 0-shot). Source: github.com. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 (Medium) | 88.14 | #16 (Claude Opus 5) |
| 2 | GPT-5.6 Sol (Medium) | 84.29 | #22 (GPT-5.6 Sol) |
| 3 | GPT-5.6 Luna (Medium) | 77 | #78 (GPT-5.6 Luna) |
| 4 | GPT-5.6 Terra (Medium) | 76.57 | #41 (GPT-5.6 Terra) |
| 5 | Claude Sonnet 5 (Medium) | 76.14 | #66 (Claude Sonnet 5) |
| 6 | DeepSeek V4 Flash (0731) (Max) | 73.43 | #85 (DeepSeek V4 Flash (0731)) |
| 7 | DeepSeek V4 Flash (0731) (High) | 73.29 | #85 (DeepSeek V4 Flash (0731)) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=korean-llm-eval-kobalt-700 · How It Works · Data refreshed daily, snapshot 2026-11-06.