Nejumi 4 - GLP - Syntactic Analysis: leaderboard
Grammaticality tasks from the jaster harness. On the JCoLA in-domain and out-of-domain splits the model labels a single Japanese sentence as grammatical or not, and on JBLiMP it picks the grammatical sentence from a pair. Each dataset is the mean of its 0-shot and 2-shot runs graded by exact match, and the board averages them in percent where higher is better. Source: github.com/wandb/llm-leaderboard.
Metric: Score (%). Source: nejumi.ai. Saturation forecast: Estimated already saturated. 139 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 (2026-04-23) | 90 | #22 |
| 2 | Gemini 3 Flash (Preview) | 89.83 | #82 |
| 3 | Gemini 3 Pro (Preview) | 89.17 | #66 |
| 4 | Gemma 4 31B (IT) | 88.83 | #176 |
| 5 | Claude Fable 5 | 88.67 | #16 |
| 6 | GPT-5.1 (2025-11-13) | 88.33 | #101 |
| 7 | GPT-5.4 (2026-03-05) | 88.17 | #59 |
| 8 | GPT-5.6 Sol | 88 | #19 |
| 9 | GPT-5.6 Terra | 88 | #41 |
| 10 | Gemini 3.5 Flash | 87.83 | #60 |
| 11 | GPT-5.6 Luna | 87.83 | #68 |
| 12 | DeepSeek V4.1 Flash | 87.67 | #36 |
| 13 | Qwen3.8 2.4T A95B | 87.67 | #55 |
| 14 | Gemini 2.5 Pro | 87.5 | #152 |
| 15 | Claude Opus 4.6 | 87.5 | #63 |
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-glp-syntactic-analysis · How It Works · Data refreshed daily, snapshot 2026-10-20.