Perplexity Sonar: Models, API Setup, Pricing and SEO Use Cases
Perplexity Sonar combines live web search with AI-generated answers and source citations. This guide compares every Sonar model, explains the API response, and shows how to control search quality, cost and output.
For the wider developer platform, see the complete Perplexity API guide.
What is Perplexity Sonar?
Perplexity Sonar is a family of web-grounded AI models available through an API. Each request can search the live web, generate an answer informed by the retrieved material and return citations plus structured search-result metadata.
It is built for developers who need current answers inside a website, application or internal workflow. Sonar is not the same as the consumer Perplexity AI interface, and it is different from the Search API, which returns raw ranked results without generating a narrative answer.
Perplexity Sonar models compared
The four models share web grounding but differ in search depth, reasoning behaviour, response detail and billing structure.
| Model | Core strength | Reasoning | Context | Input / 1M | Output / 1M | Best for |
|---|---|---|---|---|---|---|
sonar | Fast web-grounded answers | No | 128K | $1 | $1 | Quick lookups and straightforward Q&A |
sonar-pro | Deeper results and complex Q&A | No | 128K | $3 | $15 | Research requiring broader source coverage |
sonar-reasoning-pro | Multi-step analysis | Yes | 128K | $2 | $8 | Comparisons, decisions and complex reasoning |
sonar-deep-research | Exhaustive research reports | Yes | 128K | $2 | $8 | Due diligence and comprehensive analysis |
sonar-reasoning was deprecated on 15 December 2025. Use sonar-reasoning-pro for current reasoning workloads.Which Sonar model should you use?
Begin with the least expensive model that can meet the quality requirement, then evaluate it against a representative set of real questions.
Choose Sonar
Use it for factual lookups, brief current summaries, lightweight monitoring and high-volume questions that do not require elaborate reasoning.
Best starting point for cost-sensitive applicationsChoose Sonar Pro
Use it when source coverage and answer depth matter more than minimum cost. Perplexity says it returns twice as many search results as standard Sonar.
Best for deeper cited researchChoose Reasoning Pro
Use it for multi-step comparisons, complex problem solving and questions where the answer depends on combining evidence logically.
Best for reasoning over current informationChoose Deep Research
Use it selectively for exhaustive reports, due diligence and market intelligence that justify longer latency and less predictable cost.
Best for comprehensive investigationsHow the Sonar API works
Sonar searches before writing its final answer. That sequence makes the user message and formal search parameters especially important.
Receive the question
Your application sends messages, a model name and optional search controls.
Shape the search
Sonar uses the user message—not the system message—to determine retrieval.
Retrieve sources
The service searches the web under the requested domain, date, language or mode constraints.
Generate the answer
The selected model synthesises retrieved evidence using answer-time instructions.
Return evidence
The response can include citations, search-result records and detailed usage data.
Important Perplexity Sonar features
Sonar provides more than cited text. Its controls can make the output easier to integrate, audit and tailor to a specific product.
Citations and search results
Access cited URLs plus source records containing titles, URLs, dates and snippets. Keep them visible when evidence matters.
Streaming responses
All Sonar models can stream answer content. Search results, usage and other metadata arrive in the final chunk or chunks.
Structured outputs
Use a response schema when downstream software needs predictable fields instead of unstructured prose.
Search filters
Control domains, recency, dates, location, language, academic mode, SEC filings and search-context size.
Images and files
Send supported images or documents for analysis and optionally receive images, videos or related questions in the response.
SDK compatibility
Use Perplexity’s native Python or TypeScript SDK, or reuse OpenAI Chat Completions clients with Perplexity’s base URL.
How to call the Perplexity Sonar API
Create and securely store a Perplexity API key, then send it as a Bearer token to the canonical Sonar endpoint.
curl https://api.perplexity.ai/v1/sonar \
-H "Authorization: Bearer $PERPLEXITY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "sonar-pro",
"messages": [{
"role": "user",
"content": "Summarise major UK search updates announced in the last month."
}],
"search_recency_filter": "month",
"search_domain_filter": [
"developers.google.com",
"blog.google"
],
"web_search_options": {
"search_context_size": "medium"
}
}'How to control Sonar search results
Use formal parameters for hard constraints. This is more dependable than asking for restrictions only inside a system prompt.
| Control | What it changes | Example | Useful for |
|---|---|---|---|
search_domain_filter | Allows or excludes domains and URLs; maximum 20 entries | ["gov.uk", "-reddit.com"] | Trusted-source or exclusion policies |
search_recency_filter | Restricts results to a recent period | "day", "week", "month", "year" | News and update monitoring |
| Date filters | Sets publication or last-updated boundaries | After/before dates | Historical or reporting windows |
| Location and language | Targets geographic or language relevance | Country, region or language codes | Local and international research |
search_mode | Selects web or academic retrieval | "academic" | Research papers and evidence reviews |
search_context_size | Changes retrieval depth and request cost | "low", "medium", "high" | Balancing cost with coverage |
What is Sonar Pro Search?
Pro Search extends Sonar Pro with multi-step tool use for complex questions. It can search repeatedly and fetch full URL content as the research path develops.
Choose how Sonar Pro searches
The search_type parameter sits inside web_search_options. Streaming is required for Pro Search.
fastStandard Sonar Pro behaviour for direct lookups. This is the default when search_type is omitted.proForces multi-step tool usage when the question clearly requires complex research.autoLets a classifier route straightforward questions to fast and complex questions to pro.Perplexity Sonar pricing
Standard Sonar, Pro and Reasoning Pro combine token charges with a request fee based on search-context size. Deep Research has additional billing components.
Sonar
$1 input / $1 outputPer one million tokens. Request fees are $5, $8 or $12 per 1,000 for low, medium or high context.
Sonar Pro
$3 input / $15 outputRequest fees are $6, $10 or $14 per 1,000 for low, medium or high context.
Reasoning Pro
$2 input / $8 outputRequest fees are $6, $10 or $14 per 1,000 for low, medium or high context.
Deep Research
$2 input / $8 outputAlso $2 per million citation tokens, $3 per million reasoning tokens and $5 per 1,000 model searches.
Perplexity Sonar use cases for SEO and guest posting
Sonar is most useful when a task needs both current web evidence and a synthesised answer. It does not replace crawling, analytics or editorial judgement.
Search update monitoring
Track official announcements and recent commentary using recency and domain filters, then store citations beside each summary.
Publisher research
Summarise a publication’s public focus, recent coverage and visible editorial signals before a human reviews suitability.
Evidence-led topic research
Compare sources around a topic, identify recurring questions and collect dated evidence for an editorial brief.
GEO source analysis
Study which sources appear around an entity or question, while treating a small prompt sample as directional rather than definitive.
Useful Sonar prompt patterns
Keep the user message specific about the question, geography, time period and required evidence. Apply hard source restrictions through request parameters.
Update monitoring
“Summarise confirmed Google Search documentation and ranking-system updates announced during the last 30 days. Separate official changes from industry interpretation.”
Publisher research
“Analyse the recent topics, intended audience and recurring formats on this publication. Cite examples and flag anything that cannot be confirmed publicly.”
Evidence comparison
“Compare current guidance from three authoritative sources on [topic]. Show agreements, disagreements, dates and the evidence supporting each position.”
Content brief
“Create a factual research brief for [audience] about [topic]. Include key questions, recent statistics, source dates and claims requiring editorial verification.”
What Sonar cannot guarantee
Web grounding improves traceability, but it does not make every retrieved source authoritative or every generated claim correct.
Source quality still varies
A citation proves that a source was used; it does not prove the source is accurate, independent or suitable for a consequential claim.
Freshness is not completeness
A recent answer may omit inaccessible pages, paywalled evidence or sources that the retrieval process did not surface.
Automation needs boundaries
Do not let a generated answer publish content, contact publishers or make legal, medical or financial decisions without appropriate review.
Perplexity Sonar questions
Direct answers about models, live search, citations, compatibility, subscriptions and source controls.
Is Perplexity Sonar an AI model or an API?
Sonar is a family of Perplexity models delivered through the Sonar API. Applications select a model such as sonar or sonar-pro when making a request.
Does Sonar search the live web?
Yes. Sonar is designed for web-grounded responses and can return citations plus structured search-result metadata. Search behaviour can be adjusted with filters and context settings.
What is the difference between Sonar and Sonar Pro?
Standard Sonar prioritises quick, economical answers. Sonar Pro is designed for more complex questions, enhanced result accuracy and broader search coverage, but its output tokens and request fees cost more.
Can I use the OpenAI SDK with Sonar?
Yes. Configure the base URL as https://api.perplexity.ai and use a Perplexity API key. The canonical endpoint is POST /v1/sonar, while /chat/completions is accepted as a compatibility alias.
Does Sonar always provide accurate citations?
No AI retrieval system can guarantee that every source is authoritative or that every claim perfectly reflects it. Show citations, inspect important sources and add human verification for consequential output.
Do I need Perplexity Pro to use Sonar?
No consumer subscription is required. Sonar API access uses a separate API group, key, prepaid credits and pay-as-you-go pricing.
Can Sonar search only selected websites?
Yes. The search_domain_filter parameter supports allowlists and denylists with up to 20 domains or URLs. A minus prefix excludes an entry.
Use Sonar when an answer needs current web evidence
Select the lightest suitable model, constrain retrieval with parameters, preserve source metadata and review important claims before your application publishes or acts.
