Nejumi 4 - MT-Bench (Japanese) - Humanities: leaderboard
Metric: Judge rating (1-10, x10). Source: nejumi.ai. 137 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 100 |
| 2 | Qwen 3 235B A22B 2507 (Thinking) | 100 |
| 3 | Gemini 3.5 Flash | 100 |
| 4 | Gemini 3 Flash (Preview) | 100 |
| 5 | GPT-5 (High) | 100 |
| 6 | Gemini 3.6 Flash | 100 |
| 7 | GPT-5.1 (High) | 100 |
| 8 | GLM-5.2 (Max) | 100 |
| 9 | GPT-5.6 Terra (Max) | 100 |
| 10 | GLM-5.3 (Max) | 100 |
| 11 | Grok 4.20 0309 (Reasoning) | 100 |
| 12 | Claude Opus 5 (Max) | 100 |
| 13 | DeepSeek V4 Flash (Max) | 100 |
| 14 | llm-jp-4-32B-a3B (Thinking) | 100 |
| 15 | Qwen 3.5 27B (Thinking) | 100 |
Interactive version: theaggregate.ai/benchmark?slug=nejumi-4-mt-bench-japanese-humanities · How It Works · Data refreshed daily, snapshot 2026-09-19.