Horangi 4 - HRM8K: leaderboard
HRM8K, Korean math reasoning problems drawn from Korean exams and translated benchmarks, as sampled by the Horangi Korean LLM Leaderboard 4 (W&B Korea). Accuracy (%) of final answers, higher is better.
Metric: Accuracy (%). Source: horangi.ai. Saturation forecast: Estimated already saturated. 105 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 (2025-11-13) | 97 | #107 |
| 2 | Qwen 3.7 Max | 97 | #85 |
| 3 | Claude Fable 5 (High) | 97 | #20 (Claude Fable 5) |
| 4 | Gemini 3 Flash (Preview) | 96 | #90 |
| 5 | Gemini 3.5 Flash | 96 | #66 |
| 6 | Grok 4.20 | 96 | #179 |
| 7 | Claude Opus 4.6 (High) | 96 | #67 (Claude Opus 4.6) |
| 8 | Grok 4 Fast (Reasoning) | 96 | #266 (Grok 4 Fast) |
| 9 | Gemini 3 Pro (Preview) (High) | 96 | #71 (Gemini 3 Pro (Preview)) |
| 10 | GPT-5.4 Mini (2026-03-17) (xHigh) | 96 | #180 (GPT-5.4 Mini (2026-03-17)) |
| 11 | GPT-5.4 Nano (2026-03-17) (xHigh) | 96 | #285 (GPT-5.4 Nano (2026-03-17)) |
| 12 | Qwen 3.6 Plus (Preview) | 95.96 | |
| 13 | DeepSeek V3.2 (High) | 95.79 | #210 (DeepSeek V3.2) |
| 14 | Kimi K2.5 | 95 | #152 |
| 15 | Gemini 3.1 Pro (Preview) (High) | 95 | #64 (Gemini 3.1 Pro (Preview)) |
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-hrm8k · How It Works · Data refreshed daily, snapshot 2026-11-05.