Horangi 4 - KoBALT-700 (Semantics): leaderboard
KoBALT-700 semantics questions, expert-written Korean linguistic-competence items on meaning, as run by the Horangi Korean LLM Leaderboard 4 (W&B Korea). Accuracy (%), higher is better.
Metric: Accuracy (%). Source: horangi.ai. Saturation forecast: Around February 2027. 105 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (High) | 91 | #64 (Gemini 3.1 Pro (Preview)) |
| 2 | Claude Fable 5 (High) | 87 | #20 (Claude Fable 5) |
| 3 | Gemini 3 Pro (Preview) (High) | 85 | #71 (Gemini 3 Pro (Preview)) |
| 4 | GPT-5.6 Sol (Max) | 84 | #22 (GPT-5.6 Sol) |
| 5 | GPT-5.6 Terra (Max) | 84 | #40 (GPT-5.6 Terra) |
| 6 | GPT-5.5 (xHigh) | 83 | #33 (GPT-5.5) |
| 7 | Gemini 3.5 Flash | 82.83 | #66 |
| 8 | GLM-5 | 82 | #149 |
| 9 | Qwen 3.7 Max | 81.25 | #85 |
| 10 | Grok 4.5 | 80.61 | #54 |
| 11 | GPT-5 | 80 | #105 |
| 12 | Gemini 3 Flash (Preview) | 80 | #90 |
| 13 | GPT-5.1 (2025-11-13) | 80 | #107 |
| 14 | GPT-5.6 Luna (Max) | 80 | #79 (GPT-5.6 Luna) |
| 15 | GPT-5.2 (2025-12-11) (xHigh) | 80 |
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-kobalt-700-semantics · How It Works · Data refreshed daily, snapshot 2026-11-05.