Sao10K: Llama 3 8B Lunaris

Sao10K's Llama 3 8B Lunaris, created on August 12, 2024, demonstrates strong performance in operational efficiency, consistently ranking among the fastest models (81st percentile) and offering highly competitive pricing (94th percentile) across various benchmarks. This makes it an economically attractive option for deployment. However, its accuracy across most benchmark categories is generally low. It struggles significantly in specialized areas such as Coding (4.0% accuracy), Mathematics (19.0% accuracy), and Reasoning (20.0% accuracy), placing it in the lower percentiles for these domains. General Knowledge (69.0% accuracy) and Ethics (68.5% accuracy) also show below-average performance. The model's ability to acknowledge uncertainty, as measured by the Hallucinations benchmark, is moderate at 52.0% accuracy, suggesting room for improvement in its self-awareness regarding fictional concepts. Its strongest performance in terms of accuracy is in Email Classification (89.0%), though this still places it in the 13th percentile. Instruction Following is a moderate area at 41.0% accuracy. Overall, Lunaris 8B excels in cost-effectiveness and speed but requires significant improvement in accuracy and logical reasoning across a broad spectrum of tasks.