Swallow - Pre-trained Japanese - JEMHopQA: leaderboard
From the Swallow LLM Leaderboard (Institute of Science Tokyo) pre-trained Japanese board, which evaluates base models without post-training. JEMHopQA open-ended multi-hop question answering, 4-shot; metric is character-level F1. Upstream publishes 0-1 scores; shown as 0-100, higher is better.
Metric: Character F1 (%). Source: swallow-llm.github.io. Saturation forecast: Estimated already saturated. 50 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | sarashina2-70B | 71.7 | #586 |
| 2 | Llama 3.1 70B | 61.6 | #588 |
| 3 | Qwen 2.5 72B | 61.1 | #512 |
| 4 | Qwen3-30B-A3B-Base | 60.1 | #597 |
| 5 | Llama 4 Scout Base | 59.5 | #708 |
| 6 | gemma-3-27B-pt | 58.2 | #505 |
| 7 | Qwen3-14B-Base | 57.9 | #628 |
| 8 | Gemma 2 9B | 57.3 | #924 |
| 9 | Qwen 2.5 14B | 56.7 | #716 |
| 10 | gemma-3-12B-pt | 56.3 | #686 |
| 11 | sarashina2.2-3B | 56.3 | #766 |
| 12 | Qwen 2.5 32B | 56.1 | #544 |
| 13 | sarashina2-13B | 55.7 | #898 |
| 14 | Gemma 2 27B | 55.3 | #803 |
| 15 | Qwen3-8B-Base | 53.7 | #626 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=swallow-pre-trained-japanese-jemhopqa · How It Works · Data refreshed daily, snapshot 2026-11-05.