Maylin-7B: benchmark results
Azazelle's DARE-TIES merge, whose published recipe uses mistralai/Mistral-7B-v0.1 as its base and SanjiWatsuki/Sonya-7B, Azazelle/Argetsu, Azazelle/Tippy-Toppy-7b as merge inputs. Its published Mistral configuration specifies 32 decoder layers. Provider: Azazelle. Access: Open.
Aggregate Rating 1435 ± 34, rank #4420 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - HellaSwag | 86.4 | Normalized accuracy (%) (10-shot) | 80.8 |
| Open LLM Leaderboard v1 - GSM8K | 63.76 | Accuracy (%) (5-shot) | 77.4 |
| Open LLM Leaderboard v1 - MMLU | 64.73 | Accuracy (%) (5-shot) | 75.6 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 60.24 | MC2 (%) (0-shot) | 75.5 |
| Open LLM Leaderboard v1 - ARC Challenge | 66.81 | Normalized accuracy (%) (25-shot) | 74.8 |
| Open LLM Leaderboard v1 - WinoGrande | 79.64 | Accuracy (%) (5-shot) | 70.3 |
Interactive version: theaggregate.ai/model?slug=maylin-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.