Cette page est affichée en anglais. Une traduction française est en cours.
KnowledgeMay 11, 2026·5 min de lecture

Perplexity Sonar API — Search-Grounded LLM in One Call

Perplexity Sonar API returns LLM answers grounded in real-time web search with citations. Tiers: sonar / sonar-pro / sonar-reasoning.

Perplexity
Perplexity · Community
Prêt pour agents

Cet actif peut être lu et installé directement par les agents

TokRepo expose une commande CLI universelle, un contrat d'installation, le metadata JSON, un plan selon l'adaptateur et le contenu raw pour aider les agents à juger l'adaptation, le risque et les prochaines actions.

Stage only · 15/100Stage only
Surface agent
Tout agent MCP/CLI
Type
Knowledge
Installation
Stage only
Confiance
Confiance : New
Point d'entrée
Asset
Commande CLI universelle
npx tokrepo install 25b2aa98-cc43-4d6c-b654-5baa3f3c9f62
contrat d'installationJSON metadataplan adaptateurcontenu raw
Introduction

Perplexity's Sonar API is a one-call alternative to building search + scrape + chunk + RAG yourself — you send a question, Perplexity searches the web in real time and returns an LLM answer with inline numbered citations to the source URLs. Three tiers: sonar (fast/cheap), sonar-pro (better answer quality, more sources), sonar-reasoning (chain-of-thought, longer think time). Best for: news Q&A, fact-checking, anywhere you need a fresh answer with sources. Works with: OpenAI-compatible client (Python, JS), curl, LangChain. Setup time: 2 minutes.

Python (openai-compatible)

from openai import OpenAI

client = OpenAI(
    base_url="https://api.perplexity.ai",
    api_key=os.environ["PPLX_API_KEY"],
)

resp = client.chat.completions.create(
    model="sonar-pro",
    messages=[{"role": "user", "content": "What are the top 3 AI funding rounds this week?"}],
)
print(resp.choices[0].message.content)
# Response includes inline citations like [1][2][3]

# Read citation URLs separately
print(resp.citations)   # ["https://...", "https://...", "https://..."]

Filter sources by domain or recency

resp = client.chat.completions.create(
    model="sonar-pro",
    messages=[{"role": "user", "content": "What's the latest Anthropic announcement?"}],
    extra_body={
        "search_domain_filter": ["anthropic.com", "techcrunch.com"],   # whitelist
        "search_recency_filter": "week",                                # day | week | month | year
        "return_images": False,
        "return_related_questions": True,
    },
)

Model tiers (May 2026)

Model Use case Cost ($/1M) Latency
sonar Quick lookups, single-source Q&A $1 in / $1 out ~1–3s
sonar-pro Production answer quality, multi-source $3 in / $15 out ~3–7s
sonar-reasoning Hard reasoning, citations + thinking $1 in / $5 out ~10–25s
sonar-reasoning-pro Top quality reasoning $2 in / $8 out ~15–40s
sonar-deep-research Long research reports with 30+ sources $2 in / $8 out + per-search ~minutes

When NOT to use Sonar

If your data is private, not on the web, or in your own corpus — use a private RAG pipeline (e.g., Tavily + your vector store). Sonar searches public web only.

FAQ

Q: Sonar vs Grok Live Search vs Tavily? A: Grok bundles search into the same model call cheaply. Sonar gives stronger answer quality and richer citations. Tavily is search-only (you bring your own LLM). Use Sonar when answer quality matters; Tavily when you need control over the LLM stage.

Q: Are citations clickable? A: Citations come back as a citations array of URLs separately from the markdown answer. Render them as numbered footnotes in your UI. Sonar's content also embeds [1], [2] inline so you can map them visually.

Q: Rate limits? A: Standard tier: ~50 RPM on sonar, ~20 RPM on sonar-pro. Higher tiers in console.perplexity.ai. For production scaling beyond, talk to Perplexity Sales — they offer dedicated capacity.

Quick Use

  1. Get PPLX_API_KEY at perplexity.ai/settings/api
  2. OpenAI(base_url='https://api.perplexity.ai', api_key=PPLX_KEY)
  3. Use model='sonar-pro' and read resp.citations for source URLs

Intro

Perplexity's Sonar API is a one-call alternative to building search + scrape + chunk + RAG yourself — you send a question, Perplexity searches the web in real time and returns an LLM answer with inline numbered citations to the source URLs. Three tiers: sonar (fast/cheap), sonar-pro (better answer quality, more sources), sonar-reasoning (chain-of-thought, longer think time). Best for: news Q&A, fact-checking, anywhere you need a fresh answer with sources. Works with: OpenAI-compatible client (Python, JS), curl, LangChain. Setup time: 2 minutes.

Python (openai-compatible)

from openai import OpenAI

client = OpenAI(
    base_url="https://api.perplexity.ai",
    api_key=os.environ["PPLX_API_KEY"],
)

resp = client.chat.completions.create(
    model="sonar-pro",
    messages=[{"role": "user", "content": "What are the top 3 AI funding rounds this week?"}],
)
print(resp.choices[0].message.content)
# Response includes inline citations like [1][2][3]

# Read citation URLs separately
print(resp.citations)   # ["https://...", "https://...", "https://..."]

Filter sources by domain or recency

resp = client.chat.completions.create(
    model="sonar-pro",
    messages=[{"role": "user", "content": "What's the latest Anthropic announcement?"}],
    extra_body={
        "search_domain_filter": ["anthropic.com", "techcrunch.com"],   # whitelist
        "search_recency_filter": "week",                                # day | week | month | year
        "return_images": False,
        "return_related_questions": True,
    },
)

Model tiers (May 2026)

Model Use case Cost ($/1M) Latency
sonar Quick lookups, single-source Q&A $1 in / $1 out ~1–3s
sonar-pro Production answer quality, multi-source $3 in / $15 out ~3–7s
sonar-reasoning Hard reasoning, citations + thinking $1 in / $5 out ~10–25s
sonar-reasoning-pro Top quality reasoning $2 in / $8 out ~15–40s
sonar-deep-research Long research reports with 30+ sources $2 in / $8 out + per-search ~minutes

When NOT to use Sonar

If your data is private, not on the web, or in your own corpus — use a private RAG pipeline (e.g., Tavily + your vector store). Sonar searches public web only.

FAQ

Q: Sonar vs Grok Live Search vs Tavily? A: Grok bundles search into the same model call cheaply. Sonar gives stronger answer quality and richer citations. Tavily is search-only (you bring your own LLM). Use Sonar when answer quality matters; Tavily when you need control over the LLM stage.

Q: Are citations clickable? A: Citations come back as a citations array of URLs separately from the markdown answer. Render them as numbered footnotes in your UI. Sonar's content also embeds [1], [2] inline so you can map them visually.

Q: Rate limits? A: Standard tier: ~50 RPM on sonar, ~20 RPM on sonar-pro. Higher tiers in console.perplexity.ai. For production scaling beyond, talk to Perplexity Sales — they offer dedicated capacity.

Source & Thanks

Built by Perplexity. Sonar API docs at docs.perplexity.ai.

Official SDK pending; OpenAI-compatible client works today.

🙏

Source et remerciements

Built by Perplexity. Sonar API docs at docs.perplexity.ai.

Official SDK pending; OpenAI-compatible client works today.

Fil de discussion

Connectez-vous pour rejoindre la discussion.
Aucun commentaire pour l'instant. Soyez le premier à partager votre avis.

Actifs similaires

Weave — Trace and Debug LLM Apps

Weave adds tracing to LLM apps with `@weave.op`. Install `weave`, call `weave.init()`, then track inputs/outputs across API calls and validation steps.

Knowledge
Agent Toolkit

Helicone Sessions — Group LLM Calls by User Conversation

Helicone Sessions group multiple LLM calls under one session ID. Trace a multi-step agent run end-to-end, see total cost, latency, conversation flow.

Knowledge
Helicone

Helicone Cache — Cut LLM Spend with Drop-In Response Caching

Helicone Cache short-circuits identical LLM requests at the proxy. Set Helicone-Cache-Enabled header, exact-match responses come back in ms at zero cost.

Knowledge
Helicone

Statewave — Memory Runtime for AI Agents (API + SDKs)

Statewave is a self-hostable memory runtime: ingest episodes, compile memories, do semantic search, and build token-bounded context bundles via REST.

Knowledge
AI Open Source
◈Accueil🔍Rechercher👤Moi