Open Japanese LLM - Xlsum JA Bleu JA — leaderboard
Metric: Score (%). Source: huggingface.co. 418 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 2 70B Chat (HF) | 78.31 |
| 2 | Llama 2 13B Base | 75.69 |
| 3 | DeepSeek-R1-Distill-Qwen-7B | 75.15 |
| 4 | Qwen 1.5 4B | 62.09 |
| 5 | Saka-1.5B | 51.76 |
| 6 | DeepSeek R1 Distill Qwen 1.5B | 43 |
| 7 | Llama 2 70B Base | 35.45 |
| 8 | Llama 3 8B Instruct | 34.1 |
| 9 | Llama 2 13B Chat Base | 33.8 |
| 10 | Qwen 2.5 Coder 7B Instruct | 19.22 |
| 11 | Qwen2.5-Math-7B-Instruct | 14.76 |
| 12 | aya-23-35B | 4.15 |
| 13 | aya-expanse-32B | 3.66 |
| 14 | Llama 3.1 70B Instruct | 3.43 |
| 15 | Llama 3.1 Nemotron 70B Instruct | 3.41 |
Interactive version: theaggregate.ai/benchmark?slug=open-japanese-llm-xlsum-ja-bleu-ja · How the rankings work · Data refreshed daily, snapshot 2026-07-22.