ConfigurableSOLAR-10.7B: benchmark results
vicgalle's ConfigurableSOLAR 10.7B, a SOLAR fine-tune using configurable safety tuning so system prompts can switch between harmless, uncensored and persona behaviors. Provider: vicgalle. Released 2024-03-10. Access: Open.
Aggregate Rating 1444 ± 17, rank #1035 of 1661 rated models, from 27 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 72.34 | MC2 (%) (0-shot) | 93.4 |
| Open LLM Leaderboard v1 - HellaSwag | 88.03 | Normalized accuracy (%) (10-shot) | 89.2 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 54.36 | Exact match, flexible extract (%) | 88.4 |
| Open LLM Leaderboard v1 - MMLU | 66.44 | Accuracy (%) (5-shot) | 87.3 |
| Open Korean LLM - KorNAT-Harmless | 66.86 | Normalized accuracy (%) | 87 |
| Open LLM Leaderboard v1 - WinoGrande | 83.03 | Accuracy (%) (5-shot) | 86.8 |
| Open LLM Leaderboard v1 - ARC Challenge | 70.39 | Normalized accuracy (%) (25-shot) | 85.8 |
| Open Korean LLM - KorNAT-CKA | 30.43 | Normalized accuracy (%) | 83.4 |
| Open Korean LLM - Ko-EQ-Bench | 32.56 | EQ-Bench score | 81.8 |
| Open LLM Leaderboard v1 - GSM8K | 63.38 | Accuracy (%) (5-shot) | 76.7 |
| Open Portuguese LLM - FaQuAD NLI | 76.04 | Macro F1 (%) | 75.4 |
| Open Korean LLM Leaderboard | 39.65 | Average Score (%) | 73.2 |
Interactive version: theaggregate.ai/model?slug=configurablesolar-10-7b · How It Works · Data refreshed daily, snapshot 2026-10-23.