Sonar
VerifiedSonar combines multimodal input with real-time search for accurate answers.
About Sonar
Sonar was designed by Perplexity as a proprietary system that fuses vision and language processing. Its large context window enables handling of extended documents paired with images without truncation. The architecture prioritizes integration with live web data to ground outputs in current information.
Strengths include reliable multimodal reasoning and reduced hallucination through search augmentation. Because it remains closed-weight, access occurs only via Perplexity's platform rather than local deployment. Typical usage covers research tasks, image-based question answering, and complex multi-turn conversations that benefit from external knowledge retrieval.
Capabilities
Benchmarks & performance
Independent evaluation scores and measured speed.
Source: Artificial Analysis
How Sonar compares
Sonar (striped bar) vs other multimodal on intelligence, speed and price.
Intelligence
Artificial Analysis Intelligence Index · Higher is better · Sonar ranks #71 of 88
Price
USD per 1M output tokens · Lower is better · Sonar ranks #41 of 155
Sources: Artificial Analysis (intelligence, speed) · OpenRouter (price).
Best for
Real-time Cited Research
Sonar excels at answering questions that require current information by combining web search with cited outputs, making it suitable for news summaries or fact-checking tasks.
Multimodal Document Analysis
It handles long documents containing both text and images, supporting detailed reasoning over mixed inputs up to its full context length.
Extended Conversational Research
The model supports ongoing research dialogues with long-context retention, ideal for iterative exploration of complex topics.
Strengths & limitations
Strengths
- +Access to current information
- +Source-backed responses
- +Effective long-context handling
- +Strong research and summarization
Limitations
- –Dependent on search result quality
- –Less specialized for creative writing or code
- –Potential latency from web queries
Cost calculator
Estimate what Sonar would cost for your usage.
Based on Sonar's $1.00/1M input · $1.00/1M output. Estimate only — actual cost varies by provider and caching.
Quick start
OpenRouter's API is OpenAI-compatible — most SDKs work by just swapping the base URL. Only the model slug changes between models.
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://openrouter.ai/api/v1",
apiKey: process.env.OPENROUTER_API_KEY,
});
const completion = await client.chat.completions.create({
model: "perplexity/sonar",
messages: [{ role: "user", content: "Hello!" }],
});
console.log(completion.choices[0].message.content);Model slug: perplexity/sonar
Editor's verdict
Sonar is Perplexity's proprietary multimodal with a 127K-token context window.
On independent testing it scores 15.5 on the Artificial Analysis Intelligence Index.
At $1.00 per 1M output tokens, it is mid-priced for its class.
It is available through Perplexity's API and aggregators like OpenRouter.
Best suited to access to current information and source-backed responses.
Frequently asked questions
Sonar supports a context length of 127072 tokens for processing extended inputs and conversations.
User reviews
Real, verified reviews from the community shape this model's rating.
Loading reviews…
Other Sonar models
Sibling versions in the Sonar family from Perplexity.
Similar models
Other multimodal worth comparing.