MiquMaid-v2-70B: benchmark results
A Miqu-70B fine-tune for role-play conversation by NeverSleep, credited to Undi and IkariDev. The card documents Alpaca prompting and a mixture including Aesir, NoRobots and LimaRP; the plain version remains separate from DPO and mixture-of-experts releases. Provider: NeverSleep. Access: Open.
Aggregate Rating 1490 ± 36, rank #2506 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - MMLU | 75.18 | Accuracy (%) (5-shot) | 96.6 |
| Open LLM Leaderboard v1 - WinoGrande | 84.77 | Accuracy (%) (5-shot) | 95 |
| Open LLM Leaderboard v1 - HellaSwag | 87.49 | Normalized accuracy (%) (10-shot) | 86.5 |
| Open LLM Leaderboard v1 - ARC Challenge | 70.48 | Normalized accuracy (%) (25-shot) | 86.1 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 57.62 | MC2 (%) (0-shot) | 70.5 |
| Open LLM Leaderboard v1 - GSM8K | 15.62 | Accuracy (%) (5-shot) | 37.5 |
Interactive version: theaggregate.ai/model?slug=miqumaid-v2-70b · How It Works · Data refreshed daily, snapshot 2026-11-08.