Open Japanese LLM - Xlsum JA Bert Score JA F1: leaderboard
The "Xlsum JA Bert Score JA F1" 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 | Llama-3.1-Tulu-3-70B | 76.74 | #482 |
| 2 | aya-23-35B | 72.12 | #883 |
| 3 | aya-expanse-32B | 71.67 | #692 |
| 4 | Gemma 2 27B (IT) | 71.62 | #645 |
| 5 | Qwen 1.5 32B | 71.47 | #794 |
| 6 | Gemma 2 9B (IT) | 71.14 | #788 |
| 7 | Qwen 2 57B A14B | 70.97 | #783 |
| 8 | Qwen 2.5 32B | 70.91 | #546 |
| 9 | qwen2.5-bakeneko-32B-instruct-v2 | 70.76 | #562 |
| 10 | aya-23-8B | 70.72 | #1299 |
| 11 | Phi-3.5-mini-instruct | 70.67 | #1167 |
| 12 | Qwentile2.5-32B-Instruct | 70.6 | #339 |
| 13 | Qwen 2.5 14B | 70.4 | #725 |
| 14 | Qwen 1.5 32B Chat | 70.34 | #816 |
| 15 | Falcon3-10B-Instruct | 70.34 | #1157 |
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-xlsum-ja-bert-score-ja-f1 · How It Works · Data refreshed daily, snapshot 2026-10-20.