Nejumi 4 - MT-Bench (Japanese): leaderboard
Nejumi 4’s Japanese MT-Bench: multi-turn open-ended prompts graded by an LLM judge on the MT-Bench rubric, translated and re-authored for Japanese rather than machine translated.
Metric: Score (%). Source: nejumi.ai. Saturation forecast: Estimated already saturated. 139 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 98.81 | #12 |
| 2 | GLM-5.3 Flash | 98.75 | #65 |
| 3 | Inkling Small | 98.75 | #138 |
| 4 | Claude Sonnet 4.6 | 98.44 | #90 |
| 5 | Inkling | 98.38 | #126 |
| 6 | Muse Spark Open | 98.38 | #54 |
| 7 | Kimi K3 | 98.37 | #30 |
| 8 | Gemini 3.6 Flash | 98.25 | #51 |
| 9 | Claude Opus 4.8 | 98.06 | #37 |
| 10 | Claude Opus 4.7 | 98 | #49 |
| 11 | GLM-5.3 | 97.94 | #39 |
| 12 | Gemini 3.5 Flash | 97.75 | #60 |
| 13 | Claude Opus 4.6 | 97.62 | #63 |
| 14 | Gemma 4 31B (IT) | 97.62 | #176 |
| 15 | MiniMax-M3 | 97.62 | #136 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=nejumi-4-mt-bench-japanese · How It Works · Data refreshed daily, snapshot 2026-10-20.