Models·Perplexity

Perplexity Sonar: What It Does and What It Costs

perplexity/sonar

Sonar is Perplexity’s text-answer model, with support for image input.

$1Input, per million tokens
$1Output, per million tokens
127KContext window
114KLongest answer

127K tokens of context — about 95,304 words, on a scale that ends at 2M.

What it can do

  • Vision
  • Takes text
  • Takes image
  • Gives text

Released 27 Jan 2025 · checked 3 Oct 2026.

About it

Sonar is a Perplexity model for question-and-answer features. It accepts text and images and produces text. It has a 127,072-token context window and can generate up to 114,364 tokens in one answer. Input and output each cost $1 per million tokens; a typical chat turn costs about 0.15 cents.

Review

In a line

Sonar is a Perplexity model that answers questions in text. It can take text or images as input.

What it's good at

Sonar suits lightweight question-and-answer features for companies. It handles text and image input and returns text.

Price in plain terms

Input and output each cost $1 per million tokens. That works out to about 0.065 cents to read a 500-word page, about 0.15 cents for a chat turn with 1,000 tokens in and 500 out, or about $1.50 for 1,000 such turns.

Context and limits

Sonar can take in up to 127,072 tokens at once and produce up to 114,364 tokens in one answer. Its listed output is text; input can be text or images.

Ways to use it

Companies can add Sonar to question-and-answer features.

Verdict

Sonar is a fit for companies adding question-and-answer functions that need text answers and can benefit from image input. Its per-token pricing is easy to estimate for usage such as pages or chat turns. Choose another model if you need it to produce images or if your work requires outputs other than text.

Verdict

Sonar may suit a company building a question-and-answer feature that accepts text or images and responds in text. Its listed per-token rates make small workloads inexpensive by the provided examples, though total cost depends on how much the model reads and writes. Its large context and output limits allow long inputs and answers, but the model's listed output is text only. Compare it with other language models if those limits or the text-only output do not fit your needs.

  • Accepts both text and image input.
  • A 127,072-token context window supports long inputs.
  • Input and output each cost $1 per million tokens.
  • About $1.50 for 1,000 typical chat turns.
  • Produces text only.
  • Its intended use is lightweight question-and-answer features.
  • Costs rise with the number of tokens processed.
  • Pick it ifCompanies adding text-based question-and-answer features; Teams that need image input and text answers
  • Skip it ifYou need the model to generate images or other non-text output.; Your use case is not a lightweight question-and-answer feature.

Pricing

Input and output each cost $1 per million tokens. The provided estimates are about 0.065 cents to read a page, 0.15 cents per typical chat turn, and $1.50 for 1,000 such turns.

Questions

What is Sonar?

Sonar is a Perplexity model for question-and-answer features. It accepts text and images and produces text.

How much does Sonar cost?

Input and output each cost $1 per million tokens. A typical chat turn is estimated at about 0.15 cents.

How much can Sonar read at once?

Its context window is 127,072 tokens.

How long can a Sonar answer be?

It can produce up to 114,364 tokens in one answer.

Can Sonar work with images?

Yes. It accepts images as input, but its output is text.

How can a company use Sonar?

Companies can add Sonar to question-and-answer features.

On these boards

More from Perplexity