DarkSapling-7B-v2.0: benchmark results
A four-model roleplay merge released by TeeZee combining Dolphin-2.6-Mistral-DPO-Laser, Mistral-Holodeck, Mistral-Erebus and Samantha-Mistral. Version 2.0 explicitly uses DARE TIES. Provider: TeeZee. Access: Open.
Aggregate Rating 1417 ± 32, rank #5313 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - HellaSwag | 85.1 | Normalized accuracy (%) (10-shot) | 71.8 |
| Open LLM Leaderboard v1 - MMLU | 64.37 | Accuracy (%) (5-shot) | 69 |
| Open LLM Leaderboard v1 - ARC Challenge | 64.16 | Normalized accuracy (%) (25-shot) | 65.7 |
| Open LLM Leaderboard v1 - WinoGrande | 78.61 | Accuracy (%) (5-shot) | 64.3 |
| Open LLM Leaderboard v1 - GSM8K | 45.41 | Accuracy (%) (5-shot) | 59.4 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 52.21 | MC2 (%) (0-shot) | 55 |
Interactive version: theaggregate.ai/model?slug=darksapling-7b-v2-0 · How It Works · Data refreshed daily, snapshot 2026-11-08.