HomeAI models
Perplexity: Sonar
Perplexity
•
Sonar
•
Released 
January 2025

Perplexity
Sonar

Perplexity's lightweight search-grounded model for fast, citation-backed answers with adjustable web retrieval depth.

Modality:
Text
Image
model ID
perplexity/sonar
Start for free
arrow black icon
View API Docs
arrow white icon

Output Speed *

N/A
tok/s

Intelligence Index *

7.7
/ 100

Context Window *

127072
tokens

Input price

6
Anytoken

Output price

6
Anytoken
Sonar: Citation-Backed Web Answers at Lightweight Scale Sonar is Perplexity's base search model. It sits below Sonar Pro in the Sonar API lineup and above nothing in the lineup for cost-sensitive use. It combines live web retrieval with answer generation and returns citations. You can customize sources and set the search context size to low, medium or high. It has a roughly 128K-token context window. It suits quick question-answering, fact-checking and grounded lookups in high-volume applications. It is not meant for deep multi-step research. Access Sonar via API on AnyAPI.ai

Performance

Sonar's Edge Is Grounded Retrieval Without Heavy Reasoning Overhead

Sonar is built for one job: answering a question from current web sources and citing them. Perplexity positions it as the lightweight, affordable, fast option in the Sonar family, aimed at companies adding question-answering features optimized for speed. Its search context size is adjustable, so retrieval depth can be tuned per request. The practical consequence is a predictable single-pass search-and-answer step. It suits high-volume lookups. Complex analysis should go to Sonar Pro or the reasoning variants.

Technical Specifications

What the model supports

Sonar takes chat-style messages and returns text answers with citations. Its context window is roughly 128K tokens, which is enough for long prompts. Retrieved web content is injected into the prompt, so usage reports include search context as well as your own text. Retrieval depth is set with search_context_size (low, medium or high). That setting affects both answer grounding and per-request cost. Check usage.prompt_tokens instead of estimating from message length.
Verified Specifications — 
Sonar
*
Input modalities
Text
Image
output modalities
Text
Context window
127072
 tokens
Maximum output tokens
114364
Reasoning
No
Knowledge cutoff
January 2025
Pricing (standard)
6
 AnyTokens in
 / 
6
 AnyTokens out

Comparison

Sonar or Sonar Pro: When Is the Cheaper Search Tier Enough?

Sonar and Sonar Pro are Perplexity's two core search-grounded answer models. Both pair live web retrieval with citations through the same API. They differ in depth. Sonar Pro lists a 200K-token context window and returns about twice as many search results as standard Sonar, per Perplexity's model pages. Sonar lists 128K and sits in the lower-cost tier. The decision is whether your questions need broad source coverage and longer context or only a quick grounded answer.

Dimension
Sonar
Sonar Pro
Context window *
127072
tokens
200000
tokens
Output speed *
N/A
tok/s
0.00
tok/s
Intelligence Index *
7.7
7.6
Input pricing
6
AnyToken
18
AnyToken
Output pricing
6
AnyToken
90
AnyToken
Knowledge cutoff *
January 2025
January 2025

Choose Sonar when queries are simple, volume is high, latency and token cost matter more than depth, and a single-pass cited answer is enough. Choose Sonar Pro when questions are multi-part, answers need more sources, prompts or conversation history approach the 128K limit, or accuracy on complex queries outweighs the higher token rates. Sonar Pro has materially higher token pricing, so route only the harder queries to it.

Limitations & Trade-offs

Where Sonar falls short

1
Smaller context than Sonar Pro. Sonar lists about 128K tokens, while Sonar Pro lists 200K. Long documents, extended conversation histories and large injected search context compete for the same window. Perplexity's API can truncate older turns in multi-turn chats or reject oversized single messages. For long-context grounded analysis, Sonar Pro is the safer choice.
2
Search cost is not just tokens. Sonar adds request fees that vary with search context size, and retrieved web content counts toward prompt tokens. A high search context setting raises both consumption and per-request cost. Budgeting from message length alone underestimates usage, so high-volume deployments should track returned usage fields and default to low or medium settings.
3
Shallower retrieval than higher tiers. Sonar Pro returns about twice as many search results as standard Sonar, and Sonar Deep Research runs autonomous multi-search loops. Sonar answers in a single search-and-answer pass. That works for direct lookups but is weaker for comparative research, long reports or questions needing many corroborating sources. Use Sonar Pro or Deep Research for those.
4
No dedicated reasoning mode. Perplexity sells reasoning separately as Sonar Reasoning and Sonar Reasoning Pro, which use chain-of-thought. Base Sonar is the fast, direct option, so multi-step math, planning or analytical synthesis over search results are better served by those variants. This page also cites no independent benchmark scores for Sonar, so test it on your own query set before committing.

Best-Fit Workloads

Where this model earns its place

01

Fast cited Q&A


Sonar is described by Perplexity as a lightweight model for question-and-answer features optimized for speed, and it returns citations. That fits in-app help, knowledge lookups and assistant features where answers should link to current sources. The limitation is depth: it uses fewer search results than Sonar Pro, so complex queries should be escalated.

02

Fact-checking and source verification


Because answers are grounded in retrieved web content with returned citations, Sonar suits claim checking, entity lookups and verification steps that need traceable sources. Source customization lets teams narrow what is searched. Reviewers should still read the cited pages for high-stakes claims, since citations show provenance but do not guarantee correctness.

03

Time-sensitive information retrieval


Sonar searches the live web at request time, so it fits questions about recent events, current company or product details and other facts that static models may miss. Retrieval depth is adjustable, so simple lookups can use a low search context size. For extended multi-source reports, step up to Sonar Pro or Deep Research.

04

High-volume routing tier


As the lower-cost, lighter tier of the Sonar family, Sonar works as the default search model in a tiered setup. Send routine lookups to Sonar and escalate multi-part or long-context queries to Sonar Pro. The trade-off is that you need your own logic to decide which queries need escalation.

Pricing in anytokens via AnyAPI
Input
6
₳
Output
6
₳
Cache write
—
₳
Cache read
—
₳
Get API access
arrow black icon
Read Docs

on this page

PerformanceTechnical SpecificationsComparisonLimitations & Trade-offsBest-Fit WorkloadsIntegrationFAQ

Integration

Access Sonar via AnyAPI.ai

Access Sonar through AnyAPI.ai using a unified API built for multi-model AI applications. Integrate Sonar without maintaining a separate provider-specific connection, and keep the flexibility to test, switch, or combine models as your application requirements evolve.

01

One API integration

Access Sonar and other AI models through the same API workflow instead of maintaining separate integrations for every provider.

02

Easy model switching

Test Sonar against alternative models or switch models as your performance, capability, or cost requirements change without rebuilding your application around another provider API.

03

Flexible for production

Use Sonar from experimentation through production while keeping your AI stack flexible as workloads, traffic, and model requirements evolve.

04

Multi-model applications

Use Sonar for the workloads where it performs best and combine it with other models for tasks that require different capabilities, performance, or efficiency.

Start free
arrow black icon
Read Docs

Frequently Asked Questions

Answers to common questions about integrating and using this AI model via AnyAPI.ai

Sonar has a context window of roughly 128K tokens. OpenRouter lists it as 127,072 tokens, and Perplexity's documentation lists 128K. Sonar Pro is the larger tier at 200K. Retrieved web content counts toward the prompt, so usable space for your own text is smaller than the headline figure.

Sonar API pricing combines token charges with a per-request fee that varies by search context size (low, medium or high). It sits in the lower-cost tier of the Sonar family, below Sonar Pro. Because retrieved web content counts toward prompt tokens, check the usage fields in each response. See the AnyAPI.ai pricing component for current rates.

Yes. Perplexity describes Sonar as featuring citations and the ability to customize sources. Answers are grounded in retrieved web content, and the source links are returned with the response. This makes Sonar suitable for fact-checking and research assistance, where traceable sources matter more than raw generation.

Sonar Pro is the higher tier. It lists a 200K-token context window, against about 128K for Sonar, and returns roughly twice as many search results. Sonar is lighter, cheaper and aimed at fast question-answering. Choose Sonar for simple, high-volume lookups and Sonar Pro for complex, multi-part queries.

No. Base Sonar is a lightweight search-and-answer model without a dedicated reasoning mode. Perplexity offers reasoning separately as Sonar Reasoning and Sonar Reasoning Pro, which use chain-of-thought. Use Sonar when you want fast grounded answers, and a reasoning variant for multi-step problem solving over search results.

* Benchmark data source: Artificial Analysis artificialanalysis.ai