All Perplexity models
P

Sonar Deep Research

Perplexity · released March 2025

Sonar Deep Research is a Perplexity model with a 128K token context window — about 180 pages of text. It reads Text and replies in Text. It can show its reasoning before it answers.

$2.30

input / 1M tokens

$9.20

output / 1M tokens

128K

context — about 180 pages

1.9¢

a client proposal

Pricing

Input
$2.30 per million tokens
Output
$9.20 per million tokens
Web search
0.57¢ per search

Billed from your Alyph wallet. Prices can change when the provider changes theirs — last checked August 9, 2026.

The numbers

Provider
Perplexity
Released
March 2025 (2025-03-07)
Context window
128K tokens — about 180 pages
Max output
Not specified
Input
Text
Output
Text
Reasoning
Shows its thinking before answering

Cost vs Token Amount

Drag the slider to adjust the split between input and output workload.

80% Input20% Output
free$1.84$3.680 tokens250k500k750k1000k
Sonar Deep Research

Adjust the ratio slider to change how output tokens influence the final price for equivalent workloads.

What a task costs

2K tokens in · 500 tokens out.

Sonar Deep Research

Perplexity

0.92¢

for this task

50% input · 50% output

Full pricing →

Your $5 welcome credit covers about 543 of these.

Compare Sonar Deep Research

Closest in price on a typical task:

Sonar Deep Research vs GPT-4.1 →Sonar Deep Research vs Jamba Large 1.7 →Sonar Deep Research vs o3 →Sonar Deep Research vs Sonar Reasoning Pro →Any other combination →

Questions, answered

How much does Sonar Deep Research cost on Alyph?

Sonar Deep Research costs $2.30 per million input tokens and $9.20 per million output tokens on Alyph. A typical task — 5K tokens in, 800 out — works out to about 1.9¢. Billed from your Alyph wallet, with hard spending limits. Prices can change when the provider changes theirs; this page was last checked on August 9, 2026.

What is Sonar Deep Research's context window?

128K tokens — about 180 pages.

Can Sonar Deep Research read images and files?

Sonar Deep Research is text-only. The canvas still serializes uploads into plain text, so you can work with files — but it cannot see images.

Does Sonar Deep Research show its reasoning?

Yes. Sonar Deep Research can expose its thinking before the final answer, so you can watch how it got there.

Test Sonar Deep Research against GPT-4.1

One prompt, both models at the same time, the better answer wins. Free to start, with $5 of credit.

Sign up free Try the demo