Yuna-7B-Merge: benchmark results
Azazelle's DARE-TIES merge, whose published recipe uses Dans-DiscountModels/Dans-07YahooAnswers-7b as its base and Azazelle/Maylin-7b, Azazelle/smol_bruin-7b, SanjiWatsuki/Kunoichi-7B as merge inputs. Its published Mistral configuration specifies 32 decoder layers. The card heading retains another label; the exact repository and its recipe define this publication, without a global name alias. Provider: Azazelle. Access: Open.
Aggregate Rating 1441 ± 35, rank #4166 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - GSM8K | 67.63 | Accuracy (%) (5-shot) | 84 |
| Open LLM Leaderboard v1 - HellaSwag | 86.84 | Normalized accuracy (%) (10-shot) | 83.2 |
| Open LLM Leaderboard v1 - MMLU | 64.86 | Accuracy (%) (5-shot) | 77.6 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 61.2 | MC2 (%) (0-shot) | 77.2 |
| Open LLM Leaderboard v1 - ARC Challenge | 67.49 | Normalized accuracy (%) (25-shot) | 77 |
| Open LLM Leaderboard v1 - WinoGrande | 80.74 | Accuracy (%) (5-shot) | 75.4 |
Interactive version: theaggregate.ai/model?slug=yuna-7b-merge · How It Works · Data refreshed daily, snapshot 2026-11-08.