Perplexity: Sonar Reasoning Pro

Perplexity: Sonar Reasoning Pro, powered by DeepSeek R1 with Chain of Thought, demonstrates exceptional reliability with a 100% success rate across all benchmarks, consistently providing usable responses. However, it tends to have longer response times, ranking in the 15th percentile for speed, and is positioned at premium pricing levels, ranking in the 5th percentile for cost. The model exhibits strong performance in complex reasoning tasks, achieving 98.0% accuracy in the Reasoning benchmark (91st percentile), and solid results in Coding (93.0% accuracy, 83rd percentile) and Mathematics (91.0% accuracy, 68th percentile). Its General Knowledge is also robust at 99.0% accuracy. A notable weakness is its Instruction Following capability, where it scored 34.3% accuracy (34th percentile), indicating challenges with multi-layered or complex directives. Similarly, its Email Classification accuracy is relatively low at 86.0% (10th percentile). While its Ethics performance is 97.0%, its percentile ranking (31st) suggests other models may perform better in this area. Overall, Sonar Reasoning Pro excels in analytical and knowledge-based tasks but struggles with precise instruction adherence and classification, all while incurring higher costs and longer processing times.