We're excited to announce major upgrades to our Sonar models, delivering superior performance at lower costs. Recent benchmarks reveal that Sonar and Sonar Pro outperform competitors, all while remaining significantly more affordable.

Benchmark testing shows Sonar outperforms competitors across key metrics while maintaining exceptional cost efficiency. Sonar Pro surpasses even the most expensive competitor models—at a significantly lower price point.

We have added three new search modes — High, Medium, and Low — for customized performance, simplified billing, and more price control. All maintain Sonar's high factual accuracy, varying only in the amount of search compute allocated based on question complexity.

Impressive results from Perplexity! Sonar Pro’s superior performance at a lower cost is a win for developers. A 90% true positive rate and <10% false discovery rate set high standards. Can’t wait to see how it compares to GPT-4 and Claude on multi-turn conversations or code generation. Anyone tested it yet?