multimaster-7B-v5: benchmark results
An ibivibiv release with a published Mixtral configuration specifying 6 local experts, 2 selected per token and 32 decoder layers. Expert labels and nominal model sizes do not establish total or active learned parameter counts. Provider: ibivibiv. Access: Open.
Aggregate Rating 1467 ± 42, rank #2935 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - WinoGrande | 86.03 | Accuracy (%) (5-shot) | 98.9 |
| Open LLM Leaderboard v1 - HellaSwag | 88.42 | Normalized accuracy (%) (10-shot) | 91.7 |
| Open LLM Leaderboard v1 - ARC Challenge | 72.18 | Normalized accuracy (%) (25-shot) | 91.2 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 70.37 | MC2 (%) (0-shot) | 90.3 |
| Open LLM Leaderboard v1 - GSM8K | 68.01 | Accuracy (%) (5-shot) | 84.9 |
| Open LLM Leaderboard v1 - MMLU | 65.06 | Accuracy (%) (5-shot) | 81.1 |
Interactive version: theaggregate.ai/model?slug=multimaster-7b-v5 · How It Works · Data refreshed daily, snapshot 2026-11-08.