Text Generation Model

Sonar Pro

Sonar Pro is a text generation model from Perplexity that combines a 200,000-token context window with live web search.

Start Building with Sonar Pro View All Models

Publisher

Perplexity

TypeText

Context Window200,000 tokens

ReleasedMarch 2025

Input$3.00/MTok

Output$15.00/MTok

Try Sonar Pro →

About Sonar Pro

Real-time web search with cited responses

Sonar Pro is a large language model developed by Perplexity, released in March 2025. It is designed around real-time web retrieval, meaning responses are grounded in current information pulled from the web rather than relying solely on a static training dataset. The model supports a 200,000-token context window and can return citations and images alongside its text responses, making it distinct from standard chat models that lack live search integration.

Sonar Pro is well-suited for tasks that require up-to-date information, such as research, news summarization, fact-checking, and answering questions about recent events. Its citation return feature allows developers to surface the sources behind each answer, which is useful in applications where traceability and accuracy matter. The model is available through MindStudio with configurable options for returning citations and images, and it produces responses up to 8,000 tokens in length.

Capabilities

What Sonar Pro supports

Live Web Search

Retrieves real-time information from the web at inference time, allowing responses to reflect current events rather than a fixed training cutoff.

Citation Return

Optionally returns source citations alongside generated text, letting developers surface the URLs that grounded each response.

Image Return

Can return relevant images from the web as part of a response when the return_images option is enabled.

Long Context Window

Supports up to 200,000 tokens of context, enabling processing of lengthy documents or extended multi-turn conversations in a single request.

Text Generation

Generates natural language responses up to 8,000 tokens, suitable for summaries, Q&A, research briefs, and structured written content.

Ready to build with Sonar Pro?

Get Started Free

Performance

Benchmark scores

Scores represent accuracy — the percentage of questions answered correctly on each test.

Benchmark	What it tests	Score
MMLU-Pro	Expert knowledge across 14 academic disciplines	75.5%
GPQA Diamond	PhD-level science questions (biology, physics, chemistry)	57.8%
MATH-500	Undergraduate and competition-level math problems	74.5%
AIME 2024	American math olympiad problems	29.0%
LiveCodeBench	Real-world coding tasks from recent competitions	27.5%
HLE	Questions that challenge frontier models across many domains	7.9%
SciCode	Scientific research coding and numerical methods	22.6%

FAQ

Common questions about Sonar Pro

What is the context window size for Sonar Pro?

Sonar Pro supports a context window of 200,000 tokens, allowing large volumes of text to be processed in a single request.

What is the maximum response length?

The model can generate responses up to 8,000 tokens in a single output.

Does Sonar Pro have a knowledge cutoff date?

Sonar Pro performs live web retrieval at inference time, so it is not strictly limited by a static training cutoff. Responses can include information from current web sources.

Can Sonar Pro return citations for its answers?

Yes. The model includes a configurable return_citations option. When enabled, responses include references to the web sources used to generate the answer.

What is the pricing for Sonar Pro on MindStudio?

Pricing details are not published in the available metadata. You can check MindStudio's pricing page or Perplexity's API documentation for current rate information.

Resources