Swallow - Japanese MT-Bench - Average: leaderboard
Metric: Judge Score (normalized, %). Source: swallow-llm.github.io. 68 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (Medium) | 84.4 |
| 2 | GPT-5 (Medium) | 84.2 |
| 3 | GPT-5 Mini (Medium) | 83 |
| 4 | O3 (2025-04-16) (Medium) | 81.8 |
| 5 | Gemma 4 31B (IT) (Thinking) | 81.5 |
| 6 | gemma-4-26B-A4B-it (Thinking) | 80.6 |
| 7 | Qwen 3.5 27B (Thinking) | 78.4 |
| 8 | O3 Mini (Medium) | 78.2 |
| 9 | Qwen 3 235B A22B 2507 Instruct | 77.9 |
| 10 | Qwen 3.5 122B A10B (Thinking) | 77.8 |
| 11 | GPT-OSS-Swallow-120B-RL-v0.1 (Thinking) | 77.2 |
| 12 | GPT-4.1 | 77.1 |
| 13 | Qwen 3.5 35B A3B (Thinking) | 76 |
| 14 | Qwen 3 Next 80B A3B (Thinking) | 75.9 |
| 15 | Gemma 4 E4B (IT) (Thinking) | 75.8 |
Interactive version: theaggregate.ai/benchmark?slug=swallow-japanese-mt-bench-average · How It Works · Data refreshed daily, snapshot 2026-09-19.