Swallow - Pre-trained English - HellaSwag: leaderboard
From the Swallow LLM Leaderboard (Institute of Science Tokyo) pre-trained English board, which evaluates base models without post-training. HellaSwag four-choice next-event prediction, 4-shot; metric is accuracy (random 25%). Upstream publishes 0-1 scores; shown as 0-100, higher is better.
Metric: Accuracy (%). Source: swallow-llm.github.io. Saturation forecast: Estimated already saturated. 50 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.1 70B | 69 | #588 |
| 2 | Llama 4 Scout Base | 68.9 | #708 |
| 3 | Qwen 2.5 72B | 68.5 | #512 |
| 4 | Gemma 2 27B | 67.5 | #803 |
| 5 | gemma-3-27B-pt | 66.7 | #505 |
| 6 | Qwen 2.5 32B | 65.6 | #544 |
| 7 | Qwen 2.5 14B | 64.2 | #716 |
| 8 | gemma-3-12B-pt | 63.7 | #686 |
| 9 | Qwen3-30B-A3B-Base | 63.1 | #597 |
| 10 | sarashina2-70B | 62.8 | #586 |
| 11 | Gemma 2 9B | 62.6 | #924 |
| 12 | Qwen3-14B-Base | 62.5 | #628 |
| 13 | Llama 3.1 8B | 60.9 | #1130 |
| 14 | Qwen 2.5 7B | 60 | #895 |
| 15 | Falcon3-10B-Base | 59.6 | #1187 |
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-english-hellaswag · How It Works · Data refreshed daily, snapshot 2026-11-05.