Horangi 4 - KoBBQ: leaderboard
KoBBQ, the Korean Bias Benchmark for Question Answering, as sampled by the Horangi Korean LLM Leaderboard 4 (W&B Korea). Accuracy (%) over ambiguous and disambiguated contexts, higher is better.
Metric: Accuracy (%). Source: horangi.ai. Saturation forecast: Estimated already saturated. 105 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 (High) | 100 | #20 (Claude Fable 5) |
| 2 | Claude Opus 4.1 (20250805) | 99 | #139 |
| 3 | Gemini 3.1 Pro (Preview) (High) | 98 | #64 (Gemini 3.1 Pro (Preview)) |
| 4 | Gemma 4 31B (IT) (Thinking) | 98 | #178 (Gemma 4 31B (IT)) |
| 5 | Claude Sonnet 5 (High) | 97 | #65 (Claude Sonnet 5) |
| 6 | GPT-5.4 Nano (2026-03-17) (xHigh) | 97 | #285 (GPT-5.4 Nano (2026-03-17)) |
| 7 | Motif 3 | 96.97 | #195 |
| 8 | GPT-5 | 96 | #105 |
| 9 | GPT-5.5 (xHigh) | 96 | #33 (GPT-5.5) |
| 10 | Gemini 3 Pro (Preview) (High) | 96 | #71 (Gemini 3 Pro (Preview)) |
| 11 | GPT-5 Nano | 95 | #439 |
| 12 | GPT-5.4 Mini (2026-03-17) (xHigh) | 95 | #180 (GPT-5.4 Mini (2026-03-17)) |
| 13 | Qwen 3.7 Max | 94.74 | #85 |
| 14 | Gemini 3 Flash (Preview) | 94 | #90 |
| 15 | GPT-OSS-120B (High) | 94 | #346 (GPT-OSS-120B) |
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-kobbq · How It Works · Data refreshed daily, snapshot 2026-11-05.