Global AI Token Comparison Hub
CN
Back to all plans

Perplexity Sonar API

Live search · 4 Sonar models

Perplexity's own Sonar search API—four models billed per token plus request fees, separate from Pro/Max subscriptions

Token API
SubscriptionToken API
Official

Core models

sonarsonar-prosonar-reasoning-prosonar-deep-research
sonar

Perplexity's base Sonar grounded search model at $1/M input/output plus Low/Med/High context request fees—the lowest-cost entry for everyday live Q&A.

sonar-pro

Advanced Sonar at $3/$15 per 1M tokens with 2× search results and Pro Search multi-step tools—the main pick for complex queries and production search APIs.

sonar-reasoning-pro

Sonar variant with enhanced chain-of-thought at $2/$8 per 1M tokens—for multi-step logical analysis needing visible reasoning plus live web data.

sonar-deep-research

Exhaustive deep research model synthesizing hundreds of sources; base $2/$8 per 1M tokens plus separate citation/search/reasoning token charges.

Plan details

sonar

Basic live search
Input
$1
Output
$1
Official
Usage
Sonar is the base grounded search model at $1/M input and output—suited to everyday live Q&A, lightweight retrieval, and cost-sensitive calls.
Models
Request fees by search context Low/Medium/High are $5/$8/$12 per 1K calls—Low is fastest/cheapest, High is deepest for research queries.
Best for
Everyday live Q&A and cost-sensitive API integrations

sonar-pro

Advanced searchRecommended
Input
$3
Output
$15
Official
Usage
Sonar Pro targets complex queries at $3/M input and $15/M output with roughly 2× Sonar's search results; supports Pro Search multi-step tool use (search_type=pro).
Models
Standard request fees Low/Med/High = $6/$10/$14 per 1K; with Pro Search enabled $14/$18/$22 per 1K—suited to heavy research APIs needing automated multi-step search and URL fetching.
Best for
Production apps with complex multi-step search needing Pro Search

sonar-reasoning-pro

Reasoning search
Input
$2
Output
$8
Official
Usage
Sonar Reasoning Pro adds chain-of-thought reasoning on top of search at $2/M input and $8/M output—replacing retired sonar-reasoning (deprecated 2025-12-15).
Models
Suited to complex problems needing visible reasoning chains and multi-step logic with live web info; responses include reasoning tokens requiring custom JSON parsing.
Best for
Complex analysis needing reasoning chains plus live search

sonar-deep-research

Deep research
Input
$2
Output
$8
Official
Usage
Sonar Deep Research targets exhaustive research—searching hundreds of sources, synthesizing expert insights, and generating detailed reports at $2/M input and $8/M output base tokens.
Models
Additional billing: citation tokens $2/M, search queries $5/1K, reasoning tokens $3/M—suited to automated deep reports, competitive analysis, and industry scan API workflows.
Best for
Automated deep reports, industry scans, and multi-source exhaustive research

Notes

  • Request fees per 1K calls: Sonar Low/Med/High = $5/$8/$12; Sonar Pro/Reasoning Pro = $6/$10/$14; Pro Search (search_type=pro) = $14/$18/$22.
  • Search API (raw web search at $5/1K, no token fees) and Embeddings API are separate product lines outside Sonar chat model billing.
  • Agent API bills at direct third-party provider token rates (see docs/agent-api/models) with separate tool fees: web_search $0.005/call, fetch_url $0.0005/call, sandbox $0.03/session—different endpoint from Sonar API.
  • Embeddings API is priced separately: pplx-embed-v1-0.6b at $0.004/M, pplx-embed-v1-4b at $0.03/M; contextualized variants $0.008–$0.05/M. All prices in USD.

Supported coding tools

Sonar APIREST APIOpenAI SDK 兼容Pro Search

Pricing and model data sourced from official vendor websites

FAQ

General·6
General
6 条