Perplexity

Sonar

Name: Sonar
Brand: Perplexity
Price: 1.35 USD
Availability: InStock

Perplexity's Sonar is a web-grounded language model built for applications that require current, cited information. Reach for it when your workflow depends on live search results rather than static training data.

Modalities

Text+Image → Text

Below release

Not established

Weights

Not established

Where it sits in the catalogue Kyma measures

Every number against every text model Kyma prices per token — a stated rule, not a chosen line-up.

Price · input#34/59

$1.35

output $1.35

median $1.009 · best $0.0494

Price · output#15/59

$1.35

per 1M tokens generated

median $3.528 · best $0.2164

Availability · 30d#17/59

98.97%

1,265 observations since 2026-05-25

median 98.1% · best 100.0%

Throughput#51/59

21.3 tok/s

probe median, one fixed prompt

median 40.0 tok/s · best 314.3 tok/s

Response time#23/59

2.00 s

probe median, to a complete answer

median 2.28 s · best 0.84 s

Context

127K

max output 4,096

published by Perplexity, not measured here

Tick above each rail is this model, below it the other 58. Dashed rule is the field median, solid is its best. Better is left; the axis stops at the 90th percentile, so a few models sit past its right edge.

Usage

How much this model is actually called here.

Rank

#42

of 104 active models

Tokens served

515.6K

all-time

Platform share

0.0%

of all tokens

Tokens · last 15 daysAug 9 → Aug 23

Two clocks, and why they disagree

Kyma measures this model twice. Both are real; they answer different questions.

Probe · every 6h · 30 days

2.00s to answer

One fixed prompt, on a schedule, to every model. Comparable, because the model is the only thing that changes.

Observations: 1,265
Answered by a substitute: 0

Real traffic · last 7 days

5.19s to answer

Your requests, at the lengths clients actually send. Not comparable between models, but it is what running this one feels like.

Requests: 87
Completed: 100%
p95: 8.9 s

The gap is prompt length, not the model degrading. Use the probe figure to choose between models, the traffic figure to budget for your own.

Pricing

Pay per token. Cached input bills at this model’s own cached rate, listed below.

$1.35 /1M input$1.35 /1M output

Hobby10 req/day · 2K in / 500 out

~$1.01/mo

Production1,000 req/day · 2K in / 500 out

~$101/mo

Scale20,000 req/day · 2K in / 500 out

~$2,025/mo

+ Estimate your workload

Requests / day1,000

Input tokens / request2,000

Output tokens / request500

Estimated monthly cost

$101

$3.38 / day on Sonar

Same workload on:

Sonar Pro$547+440%

Llama 4 Maverick$32.40-68%

Estimates use list pricing. Actual bills depend on real token counts, and every response includes its exact cost.

When to use Sonar

Updated 2026-07-29

Where this model earns its cost — and where it doesn't.

Sonar integrates real-time web search into its generation process, returning answers grounded in current sources along with citations. It accepts text and image inputs, but outputs text only. The model operates with a 127,072-token context window and caps output at 4,096 tokens.

On Kyma, you can call this model using the search alias. It supports prompt caching, which bills repeated prefixes at this model's cached input rate. Requests route through automatic failover, and you can verify exact routing and cost in the usage.cost field and X-Kyma-Model response header. Each request incurs a small per-request search fee on top of standard token billing.

The model does not support tool calling, structured outputs, or explicit reasoning traces. It is optimized for direct, cited answers rather than complex multi-step agentic workflows or long-form generation beyond its output limit.

Live Event Tracking

Fetches current developments and cites sources automatically.

Technical Research

Answers engineering questions using up-to-date documentation.

Content Drafting

Generates articles and summaries grounded in recent web data.

Visual Analysis

Processes uploaded images alongside live web queries for context.

Not ideal for: Avoid this model for agentic workflows requiring tool use, structured JSON outputs, or long-form generation exceeding 4,096 tokens.

How it compares

Against the peers people actually weigh it against.

Spec	Sonar	Sonar Pro	Llama 4 Maverick
Input /1M	$1.35	$4.05	$0.27
Output /1M	$1.35	$20.25	$1.08
Context	127K	200K	1M
Tools	No	No	Yes
Reasoning	No	No	No
Throughput	21.3 tok/s	20.5 tok/s	32.3 tok/s

Quick start

Up and running in under two minutes.

1
Create an API key
Sign up and grab a key from the dashboard — $0.50 free credit, no card required.
Get API key →

Make your first request

Drop in your key and send a chat completion — fully OpenAI-compatible.

curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sonar",
    "messages": [
      {"role": "user", "content": "Explain prompt caching in one paragraph."}
    ]
  }'

Stream responses

Add "stream": true to receive tokens as they arrive.

curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sonar",
    "stream": true,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

FAQ

Common questions about this model.

What is the context window of Sonar?

Sonar has a 127K-token context window — roughly 187 pages of text in a single request.

How much does the Sonar API cost?

$1.35 per 1M input tokens and $1.35 per 1M output tokens. No subscription; you pay only for what you use.

Does Sonar support function calling?

No — Sonar does not support tool calling. For agent workloads, choose a tools-enabled model from the catalog.

Are the weights for Sonar publicly available?

Kyma cannot say. We could not establish it from Perplexity's own material — neither a published checkpoint under Perplexity's own account, nor a statement from Perplexity that the weights are withheld. So we publish "not established" and make no claim in either direction. Treat that as Kyma declining to guess, not as evidence about Perplexity: the honest answer to a question nobody sourced is that nobody sourced it.

Is Sonar ever served below the precision its creator released it at?

Kyma cannot say, and the reason runs one step deeper: we could not even establish whether Perplexity publishes Sonar's weights, so there may be no released checkpoint to compare a route against at all. We publish "not established" rather than reading a low format as a downgrade: on several models in this catalogue a low format is exactly how the creator ships it.

How do I use Sonar?

Kyma is OpenAI-compatible: point your SDK's base URL at https://kymaapi.com/v1, use your Kyma API key, and set the model to sonar. Signing up is free and includes $0.50 of credit — no card required.

Does this model support function calling or tool use?

No, Sonar does not support tool calling or structured outputs. It is designed for direct, cited text responses.

How does billing work for the search component?

Each request incurs a small per-request search fee on top of the standard token-based pricing.

Can I use prompt caching with this model?

Yes, repeated prompt prefixes are cached and billed at this model's cached input rate.

Start with $0.50 free credit — no card required.Create account →

More models by Perplexity

Model	Context	Input	Output
Sonar Pro	200K	$4.05	$20.25