Nejumi 4 - jaster (0-shot) - JSeM: leaderboard
JSeM, the Japanese semantic test suite of entailment problems, 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: Around June 2027. 145 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.6 Max (Preview) (Thinking) | 85 | |
| 2 | Gemini 3.1 Pro (Preview) | 84 | #64 |
| 3 | Gemini 3.6 Flash | 84 | #57 |
| 4 | Gemini 3.5 Flash | 83 | #66 |
| 5 | Claude Opus 4.5 (Thinking) | 83 | #87 (Claude Opus 4.5) |
| 6 | Claude Opus 4.8 (xHigh) | 83 | #39 (Claude Opus 4.8) |
| 7 | GPT-5.6 Terra (Max) | 82 | #40 (GPT-5.6 Terra) |
| 8 | Qwen 3.6 27B (Thinking) | 82 | #182 (Qwen 3.6 27B) |
| 9 | GPT-5.4 (2026-03-05) (High) | 82 | #63 (GPT-5.4 (2026-03-05)) |
| 10 | Gemma 4 31B (IT) | 81 | #178 |
| 11 | Gemini 3 Flash (Preview) | 81 | #90 |
| 12 | GPT-5.6 Sol (Max) | 81 | #22 (GPT-5.6 Sol) |
| 13 | Kimi K3 (Max) | 81 | #35 (Kimi K3) |
| 14 | GLM-5.3 (Max) | 81 | #44 (GLM-5.3) |
| 15 | DeepSeek V4 Pro (0813) (Max) | 81 | #58 (DeepSeek V4 Pro (0813)) |
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-jsem · How It Works · Data refreshed daily, snapshot 2026-11-05.