deepseek-r1-distill-qwen2.5-bakeneko-32B: benchmark results
Provider: DeepSeek. Access: Open.
Aggregate Rating 1548 ± 29, rank #1719 of 10916 rated models, from 96 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Nejumi 4 - Toxicity - Social Norms | 98.41 | Criteria met (%) | 87.6 |
| Nejumi 4 - MT-Bench (Japanese) - Math | 99.5 | Judge rating (1-10, x10) | 71.7 |
| Nejumi 4 - jaster (0-shot) - JSQuAD | 78.75 | Character F1 (x100) | 62.8 |
| Nejumi 4 - BFCL - Live AST | 70.37 | Accuracy (%) | 59 |
| Nejumi 4 - GLP - Logical Reasoning | 88.5 | Score (%) | 55.5 |
| Nejumi 4 - BFCL - Irrelevance Detection | 83.33 | Accuracy (%) | 54.1 |
| Nejumi 4 - BFCL - Non-Live AST | 76.3 | Accuracy (%) | 54.1 |
| Nejumi 4 - BFCL - Relevance Detection | 72.22 | Accuracy (%) | 53.1 |
| Nejumi 4 - JHumanEval | 36.41 | Sandbox pass rate (%) | 46.9 |
| Nejumi 4 - jaster (0-shot) - JSICK | 79 | Exact match (%) | 42.8 |
| Nejumi 4 - jaster (0-shot) - JCoLA (in-domain) | 73 | Exact match (%) | 42.4 |
| Nejumi 4 - GLP - Information Retrieval | 77.4 | Score (%) | 41.6 |
Interactive version: theaggregate.ai/model?slug=deepseek-r1-distill-qwen2-5-bakeneko-32b · How It Works · Data refreshed daily, snapshot 2026-11-08.