Horangi 4 - Ko-HellaSwag: leaderboard
A Korean version of HellaSwag commonsense sentence-completion, as sampled by the Horangi Korean LLM Leaderboard 4 (W&B Korea). Accuracy (%), higher is better.
Metric: Accuracy (%). Source: horangi.ai. Saturation forecast: Estimated already saturated. 105 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (xHigh) | 90 | #33 (GPT-5.5) |
| 2 | Claude Opus 4.1 (20250805) | 90 | #139 |
| 3 | Claude Sonnet 4.5 (High) | 90 | #148 (Claude Sonnet 4.5) |
| 4 | GPT-5.6 Sol (Max) | 89 | #22 (GPT-5.6 Sol) |
| 5 | Gemini 3.1 Pro (Preview) (High) | 89 | #64 (Gemini 3.1 Pro (Preview)) |
| 6 | Claude Opus 4.5 (20251101) (High) | 89 | #81 (Claude Opus 4.5 (20251101)) |
| 7 | Qwen 3.7 Plus | 88 | #98 |
| 8 | Grok 4.5 | 88 | #54 |
| 9 | Claude Opus 4.6 (High) | 88 | #67 (Claude Opus 4.6) |
| 10 | GPT-5.4 (2026-03-05) (xHigh) | 88 | #63 (GPT-5.4 (2026-03-05)) |
| 11 | GLM-5.2 (High) | 88 | #80 (GLM-5.2) |
| 12 | Qwen 3.7 Max | 87.34 | #85 |
| 13 | GPT-5 | 87 | #105 |
| 14 | Qwen 3.5 27B | 87 | #192 |
| 15 | Gemini 3 Flash (Preview) | 87 | #90 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=horangi-4-ko-hellaswag · How It Works · Data refreshed daily, snapshot 2026-11-05.