Anthropic: Claude Sonnet 4.5

Claude Sonnet 4.5 demonstrates moderate speed performance, ranking in the 38th percentile across various benchmarks. Its pricing tends to be at premium levels, positioned in the 12th percentile. A standout feature is its exceptional reliability, achieving a 100% success rate across all evaluated benchmarks, indicating consistent and stable operation. The model exhibits remarkable accuracy in several key areas. It achieved perfect scores in Hallucinations (100%), General Knowledge (100%), and Ethics (100%), often being the most accurate model at its price point and speed for these categories. Its coding capabilities are strong, scoring 95.0% accuracy, placing it in the 91st percentile. Mathematics and Email Classification also show high performance at 94.0% and 99.0% accuracy, respectively. While its Reasoning accuracy is solid at 88.0%, Instruction Following is a relative area for improvement at 67.7%. Overall, Sonnet 4.5 excels in tasks requiring factual recall, ethical judgment, and complex coding, making it highly suitable for agentic workflows and specialized domains.