Ph3della5-14B: benchmark results
allknowingroger's della_linear merge of two jpacifico Chocolatine 14B DPO fine-tunes, built on the Phi-3-medium architecture. Provider: Other. Released 2024-09-06.
Aggregate Rating 1436 ± 21, rank #1075 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-EQ-Bench | 42.72 | EQ-Bench score | 92.5 |
| Open LLM Leaderboard - BBH | 63.32 | Score | 87.2 |
| Open Korean LLM - KorNAT-CKA | 31.44 | Normalized accuracy (%) | 86.8 |
| Open LLM Leaderboard - MMLU-Pro | 47.87 | Score | 86.6 |
| Open Portuguese LLM - ASSIN2 RTE | 93.13 | Macro F1 (%) | 86.6 |
| Open LLM Leaderboard - GPQA | 34.23 | Score | 85.2 |
| Open Korean LLM - Ko-IFEval | 41.38 | Strict accuracy, prompt/instruction mean (%) | 84.4 |
| Open Portuguese LLM - BLUEX | 61.75 | Accuracy (%) | 81 |
| Open Portuguese LLM - ENEM | 71.38 | Accuracy (%) | 80.2 |
| Open Portuguese LLM - OAB Exams | 51.39 | Accuracy (%) | 79.7 |
| Open Korean LLM - KorNAT-Harmless | 64.91 | Normalized accuracy (%) | 77.6 |
| Open LLM Leaderboard - MuSR | 43.86 | Score | 77.2 |
Interactive version: theaggregate.ai/model?slug=ph3della5-14b · How It Works · Data refreshed daily, snapshot 2026-10-23.