Models·Perplexity

Sonar Reasoning Pro

perplexity/sonar-reasoning-pro

A Perplexity reasoning model that accepts text and images and returns text.

$2Input, per million tokens
$8Output, per million tokens
128KContext window
115KLongest answer

128K tokens of context — about 96,000 words, on a scale that ends at 2M.

What it can do

  • Reasoning
  • Vision
  • Takes text
  • Takes image
  • Gives text

Released 7 Mar 2025 · checked 3 Oct 2026.

About it

Sonar Reasoning Pro is a Perplexity model built for reasoning, using DeepSeek R1 and Chain of Thought. It accepts text or images, returns text, and can handle a 128,000-token context. Input costs $2 per million tokens and output costs $8 per million, with pricing that includes Perplexity search.

Review

In a line

Sonar Reasoning Pro is a Perplexity reasoning model powered by DeepSeek R1. It takes text and images as input and produces text answers.

What it's good at

This model suits tasks that need step-by-step reasoning. Its reasoning capability and Chain of Thought design are central to its purpose.

It also accepts images, so it can work with visual input as well as written prompts. See other reasoning models and vision language models.

Price in plain terms

Input is $2 per million tokens. Output is $8 per million tokens. The pricing includes Perplexity search charges.

At those rates, reading a 500-word page costs about 0.13 cents. A chat turn with about 1,000 input tokens and 500 output tokens costs about 0.6 cents. One thousand such turns cost about $6.00.

Context and limits

The model can take up to 128,000 tokens in one context. That is the total material it can consider at once.

A single answer can contain up to 115,200 tokens. Its listed output is text, even though it accepts images as input.

Ways to use it

The model ID is perplexity/sonar-reasoning-pro. Use that identifier when choosing this Perplexity model in a compatible service. See Perplexity models for related options.

Verdict

Choose Sonar Reasoning Pro when detailed reasoning, image input, and a very large reading window matter. Its text-only answers and separate input and output charges are straightforward to account for.

Skip it if you need a model that produces images or if the per-million-token rates do not fit your budget.

Verdict

Sonar Reasoning Pro is aimed at readers who need a reasoning-focused model that can accept both written and visual prompts. The 128,000-token context supports long source material, while the 115,200-token output ceiling allows lengthy replies. Costs are predictable for ordinary prompts, but output is priced higher than input and Perplexity search is included in the pricing structure. It is a sensible fit for demanding analysis in a compatible Perplexity model service, not for image generation.

  • Reasoning model powered by DeepSeek R1
  • Accepts both text and image inputs
  • 128,000-token context window
  • Supports up to 115,200 output tokens
  • Clear per-million-token input and output rates
  • Returns text rather than generated images
  • Output tokens cost more than input tokens
  • Pricing includes Perplexity search charges
  • Pick it ifLong documents that need detailed reasoning; Prompts that combine images with text; Users comparing token-based model costs
  • Skip it ifYou need image generation instead of text answers; You want pricing that excludes search charges; You do not need a large context window

Pricing

Input costs $2 per million tokens and output costs $8 per million tokens. Perplexity search pricing is included. A typical 1,000-token-in, 500-token-out turn costs about 0.6 cents.

Questions

What is Sonar Reasoning Pro?

It is a Perplexity reasoning model powered by DeepSeek R1 with Chain of Thought.

Can it understand images?

Yes. Its inputs include text and images, while its output is text.

How much context can it handle?

Its context window is 128,000 tokens.

How long can one answer be?

The maximum output is 115,200 tokens.

What does it cost?

Input is $2 per million tokens and output is $8 per million tokens. Perplexity search pricing is included.

What would 1,000 chat turns cost?

About $6.00 for turns using roughly 1,000 input tokens and 500 output tokens each.

What model ID should I look for?

The model ID is perplexity/sonar-reasoning-pro.

On these boards

More from Perplexity