Perplexity API Pricing 2026

Perplexity API uses a dual pricing model: token costs plus per-request fees. Sonar costs $1.00/$1.00 per million input/output tokens plus $5-12 per 1,000 requests (depending on search context). Sonar Pro costs $3.00/$15.00 per million tokens plus $6-14 per 1,000 requests. Sonar Reasoning Pro costs $2.00/$8.00 per million tokens. Sonar Deep Research adds citation, reasoning, and search query fees on top.

Perplexity API's Sonar models have real-time web search built in. Unlike GPT-4o or Claude which need separate search tool calls, every Sonar query automatically searches the web and grounds the response in current information. This makes Perplexity ideal for applications that need up-to-date information without building your own search pipeline.

Search context size determines how much web content Perplexity retrieves and processes per query — low, medium, or high. Higher context means more web sources, better accuracy, but higher per-request fees. Sonar ranges from $5/1K requests (low context) to $12/1K requests (high context). Choose low context for simple factual queries and high context for research-heavy tasks.

Perplexity API does not offer a free tier for API access. You need to sign up for API access and pay per usage. However, Perplexity's consumer Pro subscription ($20/month) includes generous usage of Perplexity's search interface, which is separate from the API.

For low-volume applications requiring web search, Perplexity API is often cheaper than building your own search pipeline. A DIY approach requires paying for a search API (e.g., Bing or SerpAPI), an LLM API, and embedding/retrieval infrastructure. Perplexity bundles all of this. However, at high volume, the per-request fees add up — at 1 million queries/month on Sonar high context ($12/1K), that's $12,000 just in request fees.

Perplexity's search-augmented models retrieve web citations and inject them into the context window as input tokens before generating a response. These citation tokens are billed at the standard input rate but are not part of your original prompt. This overhead is documented but not prominently surfaced, and since the number of retrieved sources varies per query, total costs per call can be difficult to predict. Developers report this as the most common source of billing surprises.

Based on developer reports, Perplexity API does not expose a parameter to limit the number of search sources retrieved per query. The model determines retrieval count autonomously, meaning citation-driven token overhead is not fully controllable on a per-request basis.

Via OpenRouter, Sonar Pro costs $3/1M input and $15/1M output tokens, while the base Sonar model costs $1/1M for both. Dedicated search APIs like Linkup charge per query rather than per token and can be paired with any LLM. For search-heavy workloads where citation provenance matters less than raw cost, purpose-built search APIs may be more economical. Perplexity's advantage is the tight integration of search context into the model's response generation.

No free tier or trial credit program for the Perplexity API was found in available source data. The API operates on a pay-as-you-go basis with no minimum commitment.

Sonar is the base search-augmented model priced at $1/1M tokens for input and output via OpenRouter — suitable for high-volume, cost-sensitive applications. Sonar Pro is priced at $3/1M input and $15/1M output, offering higher-quality and more comprehensive responses. Sonar Reasoning Pro adds reasoning capabilities at $2/1M input and $8/1M output, while Sonar Deep Research targets long-horizon research synthesis at the same per-token rate as Sonar Reasoning Pro.