Perplexity

The Perplexity Sonar API provides chat completion models with built-in search capabilities, citations, and structured output support. Perplexity models retrieve information from the web in real-time, enabling up-to-date responses with source citations.

Perplexity follows OpenAI's chat completion API format - see our OpenAI documentation for the base API details.

The four model IDs below are Perplexity's current Sonar catalog. The provider forwards model IDs without a local allow-list, but it targets the Sonar chat-completions endpoint. Perplexity's separate Agent API uses vendor-qualified IDs through a different endpoint.

Sonar and the Agent API

Perplexity now classifies Sonar Chat Completions as a legacy API. It remains supported, including by this provider, but Perplexity recommends the Agent API for new projects. The closest Agent API preset mappings are Sonar to fast, Sonar Pro to low, Sonar Reasoning Pro to medium, and Sonar Deep Research to high. See Perplexity's migration guide before moving a configuration because the Agent API uses a different request and response shape.

Setup

1. Get an API key from your Perplexity Settings
2. Set the PERPLEXITY_API_KEY environment variable or specify apiKey in your config

Supported Models

Perplexity offers several specialized models optimized for different tasks:

Model	Context Length	Description	Use Case
sonar-pro	200k	Advanced search model	Long-form content, complex queries
sonar	128k	Lightweight search model	Quick searches, cost-effective responses
sonar-reasoning-pro	128k	Premier reasoning model with Chain of Thought (CoT)	Complex analyses, multi-step problem solving
sonar-deep-research	128k	Expert-level research model	Comprehensive reports, exhaustive research

Basic Configuration

providers:
  - id: perplexity:sonar-pro
    config:
      temperature: 0.7
      max_tokens: 4000

  - id: perplexity:sonar
    config:
      temperature: 0.2
      max_tokens: 1000
      search_domain_filter: ['wikipedia.org', 'nature.com'] # Only search these domains
      search_recency_filter: 'week' # Only use recent sources

Features

Search and Citations

Perplexity models automatically search the internet and cite sources. You can control this with:

• search_domain_filter: Use an allowlist of domains or a denylist with each domain prefixed by -; the two modes cannot be mixed.
• search_recency_filter: Time filter for sources ('month', 'week', 'day', 'hour')
• return_related_questions: Get follow-up question suggestions
• web_search_options.search_context_size: Control search context amount ('low', 'medium', 'high')

providers:
  - id: perplexity:sonar-pro
    config:
      search_domain_filter: ['stackoverflow.com', 'github.com']
      search_recency_filter: 'month'
      return_related_questions: true
      web_search_options:
        search_context_size: 'high'

Date Range Filters

Control search results based on publication date:

providers:
  - id: perplexity:sonar-pro
    config:
      # Date filters - format: "MM/DD/YYYY"
      search_after_date_filter: '3/1/2025'
      search_before_date_filter: '3/15/2025'

Location-Based Filtering

Localize search results by specifying user location:

providers:
  - id: perplexity:sonar
    config:
      web_search_options:
        user_location:
          latitude: 37.7749
          longitude: -122.4194
          country: 'US' # Optional: ISO country code

Structured Output

Get responses in specific formats using JSON Schema:

providers:
  - id: perplexity:sonar
    config:
      response_format:
        type: 'json_schema'
        json_schema:
          schema:
            type: 'object'
            properties:
              title: { type: 'string' }
              year: { type: 'integer' }
              summary: { type: 'string' }
            required: ['title', 'year', 'summary']

Note: First request with a new schema may take 10-30 seconds to prepare. For reasoning models, the response will include a <think> section followed by the structured output.

Image Support

Enable image retrieval in responses:

providers:
  - id: perplexity:sonar-pro
    config:
      return_images: true

Perplexity citations are exposed through the standard metadata.citations field. The raw citations, search_results, images, and related_questions arrays returned by the API are preserved under metadata.perplexity; for example, returned image results are available at metadata.perplexity.images.

Cost Tracking

promptfoo uses the total returned by Perplexity in usage.cost.total_cost, including the charges calculated by the API for that request. If the API omits a valid total, cost is unavailable; prompt and completion token counts alone omit request, search, citation, or reasoning charges. For cached responses with known cost, cost retains that value for assertions and the evaluator records incurredCost: 0. See the Sonar response schema and Perplexity pricing.

The legacy usage_tier option does not determine billing. Use the documented search controls, such as web_search_options.search_context_size, to configure request behavior.

Advanced Use Cases

Comprehensive Research

For in-depth research reports:

providers:
  - id: perplexity:sonar-deep-research
    config:
      temperature: 0.1
      max_tokens: 4000
      search_domain_filter: ['arxiv.org', 'researchgate.net', 'scholar.google.com']
      passthrough:
        reasoning_effort: 'high' # low, medium (default), or high

Step-by-Step Reasoning

For problems requiring explicit reasoning steps:

providers:
  - id: perplexity:sonar-reasoning-pro
    config:
      temperature: 0.2
      max_tokens: 3000

Offline Creative Tasks

The former offline r1-1776 model is retired. For Sonar Pro responses without web search, set disable_search through passthrough:

providers:
  - id: perplexity:sonar-pro
    config:
      passthrough:
        disable_search: true

Best Practices

Model Selection

• sonar-pro: Use for complex queries requiring detailed responses with citations
• sonar: Use for factual queries and cost efficiency
• sonar-reasoning-pro: Use for step-by-step problem solving
• sonar-deep-research: Use for comprehensive reports (may take 30+ minutes)

Search Optimization

• Set search_domain_filter to trusted domains for higher quality citations
• Use search_recency_filter for time-sensitive topics
• For Sonar, Sonar Pro, and Sonar Reasoning Pro, set web_search_options.search_context_size to "low" to reduce request fees
• For Deep Research, set passthrough.reasoning_effort to balance depth and cost

Structured Output Tips

• When using structured outputs with reasoning models, responses will include a <think> section followed by the structured output
• JSON schemas cannot include recursive structures or unconstrained objects

Example Configurations

Check our perplexity.ai-example with multiple configurations showcasing Perplexity's capabilities:

• promptfooconfig.yaml: Basic model comparison
• promptfooconfig.structured-output.yaml: JSON Schema structured output
• promptfooconfig.search-filters.yaml: Date and location-based filters
• promptfooconfig.research-reasoning.yaml: Specialized research and reasoning models

You can initialize these examples with:

npx promptfoo@latest init --example provider-perplexity

Pricing and Rate Limits

Token prices are only part of each request's cost:

Model	Input Tokens (per million)	Output Tokens (per million)
sonar	$1	$1
sonar-pro	$3	$15
sonar-reasoning-pro	$2	$8
sonar-deep-research	$2	$8

Search request fees vary by web_search_options.search_context_size:

Model	Low (per 1K requests)	Medium (per 1K requests)	High (per 1K requests)
sonar	$5	$8	$12
sonar-pro	$6	$10	$14
sonar-reasoning-pro	$6	$10	$14

Deep Research also charges $2 per million citation tokens, $5 per 1,000 search queries, and $3 per million reasoning tokens.

Check Perplexity's pricing page for current rates.

Troubleshooting

• Long Initial Requests: First request with a new schema may take 10-30 seconds
• Citation Issues: Use search_domain_filter with trusted domains for better citations
• Timeout Errors: For research models, consider increasing your request timeout settings
• Reasoning Format: For reasoning models, outputs include <think> sections, which may need parsing for structured outputs