Qwenstein2.5-32B-Instruct: benchmark results
Qwenstein2.5-32B-Instruct is a maldv Qwen 2.5-family merge using normalized denoised Fourier interpolation of the exact models listed in its release recipe. It remains distinct from its parent checkpoints and separately published quantized conversions. Provider: maldv. Access: Open.
Aggregate Rating 1568 ± 35, rank #1433 of 10916 rated models, from 50 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Japanese LLM - Jmmlu Exact Match | 76.36 | Score (%) | 97.1 |
| Open Japanese LLM - HE | 78.51 | Score (%) | 96.9 |
| Open Japanese LLM - Jamp Exact Match | 72.13 | Score (%) | 96 |
| Open Japanese LLM - Jsem Exact Match | 80.81 | Score (%) | 95.8 |
| Open Japanese LLM - MMLU EN Exact Match | 80.66 | Score (%) | 95.2 |
| Open Japanese LLM - NLI | 81.42 | Score (%) | 94.7 |
| Open Japanese LLM - MR | 94.2 | Score (%) | 94 |
| Open Japanese LLM - Jcommonsenseqa Exact Match | 96.25 | Score (%) | 93.6 |
| Open Japanese LLM - Jsts Spearman | 88 | Score (%) | 93.5 |
| Open Japanese LLM Leaderboard | 65.93 | Average Score (%) | 90.7 |
| Open Japanese LLM - Jnli Exact Match | 89.32 | Score (%) | 90.3 |
| Open Japanese LLM - ALT E TO J Comet Wmt22 | 91.07 | Score (%) | 90.1 |
Interactive version: theaggregate.ai/model?slug=qwenstein2-5-32b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.