aegolius-acadicus-34B-v3: 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 1423 ± 36, rank #5044 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 63.33 | MC2 (%) (0-shot) | 80.5 |
| Open LLM Leaderboard v1 - ARC Challenge | 67.66 | Normalized accuracy (%) (25-shot) | 77.7 |
| Open LLM Leaderboard v1 - HellaSwag | 85.54 | Normalized accuracy (%) (10-shot) | 75.2 |
| Open LLM Leaderboard v1 - GSM8K | 54.21 | Accuracy (%) (5-shot) | 65.9 |
| Open LLM Leaderboard v1 - WinoGrande | 78.69 | Accuracy (%) (5-shot) | 64.9 |
| Open LLM Leaderboard v1 - MMLU | 62.13 | Accuracy (%) (5-shot) | 54.1 |
Interactive version: theaggregate.ai/model?slug=aegolius-acadicus-34b-v3 · How It Works · Data refreshed daily, snapshot 2026-11-08.