Nejumi 4 - jaster (0-shot) - JNLI: leaderboard
JNLI (JGLUE), Japanese natural language inference, from the llm-jp-eval jaster suite as run 0-shot by Nejumi Leaderboard 4 (Weights & Biases Japan). Score is exact match (%) on Nejumi's sampled test split; higher is better.
Metric: Exact match (%). Source: nejumi.ai. Saturation forecast: Estimated already saturated. 145 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 (Thinking) | 93 | #67 (Claude Opus 4.6) |
| 2 | Claude Opus 4 (Non-reasoning) | 93 | #170 (Claude Opus 4) |
| 3 | Gemini 3.1 Pro (Preview) | 92 | #64 |
| 4 | Claude Opus 4.8 (xHigh) | 92 | #39 (Claude Opus 4.8) |
| 5 | Claude Opus 4.5 (Thinking) | 91 | #87 (Claude Opus 4.5) |
| 6 | Gemini 3 Pro (Preview) | 90 | #71 |
| 7 | Gemini 3.6 Flash | 90 | #57 |
| 8 | Kimi K2.5 (Thinking) | 90 | #152 (Kimi K2.5) |
| 9 | Claude Opus 5 (Max) | 90 | #16 (Claude Opus 5) |
| 10 | Gemini 2.5 Pro | 89 | #159 |
| 11 | Kimi K3 (Max) | 89 | #35 (Kimi K3) |
| 12 | Claude Sonnet 4.6 (Thinking) | 89 | #94 (Claude Sonnet 4.6) |
| 13 | Qwen 3.5 27B (Thinking) | 89 | #192 (Qwen 3.5 27B) |
| 14 | Qwen 3.5 397B A17B (Thinking) | 89 | #154 (Qwen 3.5 397B A17B) |
| 15 | Claude Opus 4.7 (xHigh) | 89 | #51 (Claude Opus 4.7) |
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-jaster-0-shot-jnli · How It Works · Data refreshed daily, snapshot 2026-11-05.