llm-jp-3.1-13B-instruct4: benchmark results
National Institute of Informatics' llm-jp-3.1-13b-instruct4, an instruction-tuned 13B model in the LLM-jp-3.1 series, adding mid-training (instruction pre-training) on top of the original LLM-jp-3 recipe. Provider: National Institute of Informatics. Released 2025-05-23. Access: Open.
Aggregate Rating 1487 ± 26, rank #808 of 1635 rated models, from 87 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Japanese LLM - EL | 76.52 | Score (%) | 100 |
| Open Japanese LLM - FA | 62.07 | Score (%) | 100 |
| Open Japanese LLM - Jamp Exact Match | 99.71 | Score (%) | 100 |
| Open Japanese LLM - Jnli Exact Match | 94.33 | Score (%) | 100 |
| Open Japanese LLM - Jsem Exact Match | 96.21 | Score (%) | 100 |
| Open Japanese LLM - Jsick Exact Match | 92.55 | Score (%) | 100 |
| Open Japanese LLM - Jsts Pearson | 93.75 | Score (%) | 100 |
| Open Japanese LLM - Kuci Exact Match | 92.62 | Score (%) | 100 |
| Open Japanese LLM - MC | 94.5 | Score (%) | 100 |
| Open Japanese LLM - NLI | 96.5 | Score (%) | 100 |
| Open Japanese LLM - RC | 94.48 | Score (%) | 100 |
| Open Japanese LLM - Wiki Coreference SET F1 | 40.14 | Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=llm-jp-3-1-13b-instruct4 · How It Works · Data refreshed daily, snapshot 2026-10-15.