Open Japanese LLM - FA: leaderboard
The "FA" subtask of Open Japanese LLM. A task from the Open Japanese LLM Leaderboard, evaluating models on Japanese-language knowledge, reasoning and understanding.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 418 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | llm-jp-3.1-13B-instruct4 | 62.07 | #835 |
| 2 | llm-jp-3.1-1.8B-instruct4 | 53.84 | #1273 |
| 3 | Llama-3.3-Swallow-70B-Instruct-v0.4 | 32.82 | #529 |
| 4 | Llama-3.1-Tulu-3-70B | 31.84 | #482 |
| 5 | qwen2.5-bakeneko-32B-instruct-v2 | 29.74 | #562 |
| 6 | Llama 3.1 70B | 29.59 | #592 |
| 7 | Llama 3 70B | 29.42 | #631 |
| 8 | aya-expanse-32B | 29.35 | #692 |
| 9 | Llama 3 70B Instruct | 29.22 | #634 |
| 10 | Qwentile2.5-32B-Instruct | 29.17 | #339 |
| 11 | QwQ 32B-Preview | 28.13 | #651 |
| 12 | Awqward2.5-32B-Instruct | 28.11 | #371 |
| 13 | DeepSeek-R1-Distill-Qwen-32B-Japanese | 28.03 | #638 |
| 14 | Qwen 2.5 32B | 27.89 | #546 |
| 15 | Qwen2.5-14B-Instruct-1M | 27.79 | #578 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-japanese-llm-fa · How It Works · Data refreshed daily, snapshot 2026-10-20.