Swallow - Pre-trained English - BBH: leaderboard
From the Swallow LLM Leaderboard (Institute of Science Tokyo) pre-trained English board, which evaluates base models without post-training. BIG-Bench Hard (23 tasks), 3-shot chain of thought; metric is exact match. Upstream publishes 0-1 scores; shown as 0-100, higher is better.
Metric: Exact Match (%). Source: swallow-llm.github.io. Saturation forecast: Estimated already saturated. 50 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 72B | 84.7 | #512 |
| 2 | Qwen 2.5 32B | 81.9 | #544 |
| 3 | Llama 3.1 70B | 80.7 | #588 |
| 4 | Qwen3-30B-A3B-Base | 80.4 | #597 |
| 5 | Qwen3-14B-Base | 78.7 | #628 |
| 6 | Qwen3-8B-Base | 77.5 | #626 |
| 7 | Gemma 2 27B | 76.2 | #803 |
| 8 | Falcon3-10B-Base | 76.2 | #1187 |
| 9 | Qwen 2.5 14B | 75.1 | #716 |
| 10 | Qwen3-4B-Base | 71.3 | #880 |
| 11 | gemma-3-27B-pt | 71.1 | #505 |
| 12 | Gemma 2 9B | 70.8 | #924 |
| 13 | Qwen 2.5 7B | 69 | #895 |
| 14 | gemma-3-12B-pt | 68.1 | #686 |
| 15 | Falcon3-7B-Base | 67.3 | #1286 |
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-bbh · How It Works · Data refreshed daily, snapshot 2026-11-05.