Improved Sonar models: Industry leading performance at lower costs
Blog post from Perplexity
Perplexity announced updates to its Sonar, Sonar Pro, and Sonar Reasoning Pro models aimed at improving search performance, factual accuracy, pricing transparency, and user control over compute costs. The models now offer High, Medium, and Low search modes, except for Sonar Deep Research, allowing users to choose between maximum context for complex tasks, balanced performance, or lower-cost handling of straightforward queries; Low mode matches prior Sonar pricing. The company says benchmark tests show Sonar exceeding GPT-4o Mini’s factual accuracy and approaching GPT-4o High, while Sonar Pro reportedly surpasses higher-priced competitors, including search-enabled GPT-4o, at lower costs. Citation tokens will no longer be billed for Sonar Pro and Sonar Reasoning Pro responses in the new modes, although Sonar Deep Research remains excluded. Existing billing was scheduled to remain the default for 30 days while the modes were introduced as opt-in, followed by a shift to Low mode as default, and after April 18, 2025, certain citation and search-result usage fields would no longer be returned by the Pro APIs.
No tracked trend matches for this post yet.
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.