Midnight-Miqu-70B-v1.5: benchmark results
sophosympatheia's DARE-linear merge of Midnight-Miqu v1.0 with Tess 70B, a miqu-based 70B tuned for roleplay and storytelling with 32k context. Provider: sophosympatheia. Released 2024-03-11. Access: Open.
Aggregate Rating 1488 ± 40, rank #2539 of 10916 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - BBH | 56.06 | Score | 77.1 |
| Open Chinese LLM - TruthfulQA MC | 58.1 | Accuracy (%) | 74.9 |
| Open LLM Leaderboard - IFEval | 61.18 | Score | 73.3 |
| Open Chinese LLM Leaderboard | 60.08 | Official Nonzero-Task Mean (%) | 68 |
| Open Chinese LLM - CMMLU | 59.23 | Accuracy (%) | 67.5 |
| Open Chinese LLM - HellaSwag | 62.15 | Accuracy (%) | 66.9 |
| Open LLM Leaderboard - MMLU-Pro | 38.25 | Score | 66.7 |
| Open LLM Leaderboard - MuSR | 42.44 | Score | 64.9 |
| Open Chinese LLM - C-Eval Semantic | 74.62 | Accuracy (%) | 64.5 |
| Open Chinese LLM - WinoGrande | 64.09 | Accuracy (%) | 64.1 |
| Open Chinese LLM - ARC Challenge | 54.1 | Accuracy (%) | 63.3 |
| Open LLM Leaderboard - GPQA | 29.61 | Score | 51.6 |
Interactive version: theaggregate.ai/model?slug=midnight-miqu-70b-v1-5 · How It Works · Data refreshed daily, snapshot 2026-11-08.