Perplexity adds source citation to its Sonar API for better search capabilities

Editorial Team

January 22, 2025
3 Min Read

Perplexity has upgraded its Sonar API, letting developers integrate AI-powered search capabilities – complete with source citations – into their applications. The company now offers two versions of the API, each designed for different use cases.

Unlike traditional language models that rely solely on training data, Perplexity’s approach maintains a live connection to the internet, pulling real-time information from reliable sources. Developers can specify which sources Sonar should use when searching. The company emphasizes that it doesn’t use any API-processed data to train its models.

Sonar Pro offers deeper search capabilities

Users can choose between two tiers: standard Sonar for simpler, faster queries, and Sonar Pro for more complex tasks requiring detailed responses. The Pro version runs multiple searches for each query, typically generating twice as many source citations as the basic version.

The pricing structure reflects these differences. Standard Sonar costs $5 per 1,000 searches plus $1 per 750,000 words (input and output combined). Sonar Pro maintains the same $5 per 1,000 searches but charges $3 per 750,000 input words and $15 per 750,000 generated words.

Comparative product presentation: Sonar Pro and Sonar with features, search functions and price models on a dark green background. — A side-by-side comparison of the two Sonar versions shows their respective feature sets and usage scenarios. The differences are mainly in the depth of the search and the analysis options. | Image: Perplexity

In OpenAI’s SimpleQA benchmark, which measures how accurately AI systems answer questions, Sonar Pro performed better than models from Google with search grounding. The comparison with OpenAI and Anthropic is moderately misleading because neither model had access to the Internet.

Horizontal bar chart: SimpleQA benchmark results of various AI models, with Sonar Pro leading with 85.8 points ahead of the Gemini variants. — In the SimpleQA benchmark for factual accuracy, Sonar Pro achieves the top position with 85.8 points. | Image: Perplexity

The numbers show Sonar Pro achieved an F-Score of 0.858 (with 1.0 being perfect), while the basic version scored 0.773. This score combines both precision and completeness of results.

Expanding revenue streams

The API represents another revenue stream for Perplexity, complementing its existing Pro membership subscriptions. While Perplexity previously offered an API for search queries, this new release’s source citation feature makes it more useful for many applications. However, the AI API market is becoming increasingly competitive and price sensitive.

The company secured $500 million in funding in December 2024, reaching a $9 billion valuation. They plan to use this capital for acquisitions and media partnerships to expand their data sources.

However, Perplexity isn’t without its critics. Some have pointed out that the service sometimes copies content verbatim from sources, and like all AI search engines, it can make mistakes in its responses.

Recommendation