Anthropic: Claude Sonnet 4

Anthropic's Claude Sonnet 4 demonstrates a balanced performance profile, excelling in several key areas while maintaining competitive response times. It performs among models with competitive response times, ranking in the 45th percentile for speed. However, its pricing tends to be at premium levels, placing it in the 12th percentile. A standout feature is its exceptional reliability, achieving a 100% success rate across all benchmarks, indicating minimal technical failures. Sonnet 4 exhibits remarkable accuracy in specialized tasks, achieving perfect scores in both Hallucinations (100%) and Ethics (100%), indicating strong adherence to factual integrity and ethical principles. It also shows high proficiency in General Knowledge (99.8%), Coding (93.0%), and Mathematics (93.0%), with its SWE-bench performance of 72.7% further solidifying its coding prowess. While its Instruction Following (66.0%) and Reasoning (88.0%) scores are solid, they represent areas with potential for further enhancement compared to its top-tier accuracy in other domains. Its efficiency is notable in Hallucinations, where it is the most accurate model at its price point and among models of similar speed.