ConfigurableBeagle-11B: benchmark results
vicgalle's configurable-safety-tuning fine-tune of his CarbonBeagle-11B merge, letting the system prompt set the model's safety level. Provider: vicgalle. Released 2024-02-17. Access: Open.
Aggregate Rating 1439 ± 16, rank #1056 of 1656 rated models, from 35 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 77.13 | MC2 (%) (0-shot) | 96.6 |
| Open LLM Leaderboard v1 - HellaSwag | 88.85 | Normalized accuracy (%) (10-shot) | 94.4 |
| Open Chinese LLM - TruthfulQA MC | 63.65 | Accuracy (%) | 93.9 |
| Open Korean LLM - KorNAT-Harmless | 70.39 | Normalized accuracy (%) | 93.5 |
| Open Korean LLM - KorNAT-CKA | 33.59 | Normalized accuracy (%) | 92.5 |
| Open LLM Leaderboard v1 - ARC Challenge | 72.53 | Normalized accuracy (%) (25-shot) | 92.4 |
| Open LLM Leaderboard v1 - MMLU | 66.71 | Accuracy (%) (5-shot) | 88 |
| Open LLM Leaderboard v1 - WinoGrande | 83.27 | Accuracy (%) (5-shot) | 87.9 |
| Open Portuguese LLM - ASSIN2 RTE | 92.77 | Macro F1 (%) | 83.4 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 49.2 | Exact match, flexible extract (%) | 82.8 |
| Open Korean LLM - Ko-IFEval | 40.17 | Strict accuracy, prompt/instruction mean (%) | 82.2 |
| Open Portuguese LLM - FaQuAD NLI | 77.82 | Macro F1 (%) | 81.3 |
Interactive version: theaggregate.ai/model?slug=configurablebeagle-11b · How It Works · Data refreshed daily, snapshot 2026-10-20.