Open Japanese LLM - SUM: leaderboard
The "SUM" 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 | 22.03 | #482 |
| 2 | aya-23-35B | 14.43 | #883 |
| 3 | aya-expanse-32B | 13.88 | #692 |
| 4 | Gemma 2 27B (IT) | 13.22 | #645 |
| 5 | Qwen 1.5 32B | 12.61 | #794 |
| 6 | Gemma 2 9B (IT) | 12.44 | #788 |
| 7 | Qwen 2 57B A14B | 12.17 | #783 |
| 8 | aya-23-8B | 12.13 | #1299 |
| 9 | Qwen 2.5 32B | 12.05 | #546 |
| 10 | Qwentile2.5-32B-Instruct | 11.46 | #339 |
| 11 | QwQ 32B-Preview | 11.34 | #651 |
| 12 | Qwen 2.5 14B | 11.31 | #725 |
| 13 | Falcon3-10B-Instruct | 11.28 | #1157 |
| 14 | Lamarckvergence-14B | 11.08 | #545 |
| 15 | Linkbricks-Horizon-AI-Korean-Superb-22B | 11.04 | #708 |
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-sum · How It Works · Data refreshed daily, snapshot 2026-10-20.