Korean LLM Eval - MuSR (Ko): leaderboard
daekeun-ml Korean proficiency suite on the Korean translation of MuSR (Multistep Soft Reasoning): overall accuracy % on narrative multi-step reasoning problems, 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) | 86.53 | #16 (Claude Opus 5) |
| 2 | GPT-5.6 Sol (Medium) | 85.31 | #22 (GPT-5.6 Sol) |
| 3 | GPT-5.6 Terra (Medium) | 83.15 | #40 (GPT-5.6 Terra) |
| 4 | DeepSeek V4 Flash (0731) (Max) | 82.4 | #84 (DeepSeek V4 Flash (0731)) |
| 5 | Claude Sonnet 5 (Medium) | 81.6 | #65 (Claude Sonnet 5) |
| 6 | DeepSeek V4 Flash (0731) (High) | 81.2 | #84 (DeepSeek V4 Flash (0731)) |
| 7 | GPT-5.6 Luna (Medium) | 71.47 | #79 (GPT-5.6 Luna) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=korean-llm-eval-musr-ko · How It Works · Data refreshed daily, snapshot 2026-11-05.