Horangi 4 - ALT Average: leaderboard
Alignment (ALT) average of the Horangi Korean LLM Leaderboard 4 (W&B Korea): the mean of controllability (IFEval-Ko), harm prevention (Korean hate speech), bias prevention (KoBBQ), ethics (Ko-Moral) and hallucination prevention. 0-100, higher is better.
Metric: Score (%). Source: horangi.ai. Saturation forecast: Around February 2027. 105 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (High) | 84.85 | #64 (Gemini 3.1 Pro (Preview)) |
| 2 | Qwen 3.7 Max | 81.57 | #85 |
| 3 | Claude Fable 5 (High) | 81.32 | #20 (Claude Fable 5) |
| 4 | Gemini 3 Flash (Preview) | 81.2 | #90 |
| 5 | GLM-5.2 (High) | 81.1 | #80 (GLM-5.2) |
| 6 | Gemini 3 Pro (Preview) (High) | 81.05 | #71 (Gemini 3 Pro (Preview)) |
| 7 | GPT-5.5 (xHigh) | 80.8 | #33 (GPT-5.5) |
| 8 | GLM-5-Turbo | 80.38 | #121 |
| 9 | Gemini 3.5 Flash | 80.12 | #66 |
| 10 | Claude Opus 4.1 (20250805) | 80 | #139 |
| 11 | Claude Opus 4.6 (High) | 79.97 | #67 (Claude Opus 4.6) |
| 12 | Claude Opus 4.8 (High) | 79.57 | #39 (Claude Opus 4.8) |
| 13 | Claude Opus 4.5 (20251101) (High) | 78.91 | #81 (Claude Opus 4.5 (20251101)) |
| 14 | Gemma 4 31B (IT) (Thinking) | 78.85 | #178 (Gemma 4 31B (IT)) |
| 15 | GPT-5.6 Sol (Max) | 78.6 | #22 (GPT-5.6 Sol) |
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-alt-average · How It Works · Data refreshed daily, snapshot 2026-11-05.