Nejumi 4 - Total Score: leaderboard
The headline number of Nejumi 4, the Japanese leaderboard run by Weights and Biases, computed as a weighted mean over the suite's second-level categories. Reasoning, knowledge and question answering, and application development count double, while applied language, foundational language and each of the six alignment categories count once. Higher is better. Source: github.com/wandb/llm-leaderboard.
Metric: Score (%). Source: nejumi.ai. Saturation forecast: Around January 2027. 139 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 87.2 | #12 |
| 2 | Claude Fable 5 | 86.99 | #16 |
| 3 | Qwen3.8 2.4T A95B | 85.98 | #55 |
| 4 | Claude Opus 4.8 | 85.23 | #37 |
| 5 | Claude Opus 4.7 | 85.09 | #49 |
| 6 | GPT-5.6 Sol | 84.99 | #19 |
| 7 | GPT-5.6 Terra | 84.4 | #41 |
| 8 | Kimi K3 | 84.32 | #30 |
| 9 | Gemini 3.1 Pro (Preview) | 84.3 | #58 |
| 10 | GPT-5.5 (2026-04-23) | 84.11 | #22 |
| 11 | GPT-5.4 (2026-03-05) | 83.97 | #59 |
| 12 | Claude Opus 4.6 | 83.94 | #63 |
| 13 | Muse Spark Open | 83.79 | #54 |
| 14 | Hy4 preview | 83.44 | #38 |
| 15 | Gemini 3.6 Flash | 83.12 | #51 |
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-total-score · How It Works · Data refreshed daily, snapshot 2026-10-20.