AionLabs: Aion-1.0-Mini

AionLabs: Aion-1.0-Mini, a 32B parameter distilled variant of DeepSeek-R1, demonstrates exceptional speed and competitive pricing, consistently ranking among the fastest and most cost-effective models across all evaluated benchmarks. Created on February 4, 2025, with a substantial context length of 131072, this model aims for strong performance in reasoning domains. While excelling in efficiency, its accuracy varies significantly across categories. Aion-1.0-Mini shows a notable strength in Email Classification, achieving 99.0% accuracy (81st percentile), indicating a strong ability to understand context and categorize information. However, its performance in other areas is generally lower. It exhibits a moderate ability to acknowledge uncertainty in fictional scenarios with 74.0% accuracy in Hallucinations (20th percentile). Key weaknesses are apparent in Instruction Following (0.0% accuracy), Reasoning (26.0% accuracy, 14th percentile), Ethics (41.0% accuracy, 12th percentile), and General Knowledge (49.3% accuracy, 15th percentile). Its performance in Coding (36.0% accuracy, 17th percentile) and Mathematics (51.0% accuracy, 19th percentile) also falls into the lower percentiles. This suggests that while highly efficient and cost-effective, Aion-1.0-Mini may require further development to enhance its accuracy in complex reasoning, ethical judgment, and broad knowledge domains.