# Sonar Pro

Sonar Pro extends the Sonar search architecture with deeper source retrieval and more thorough synthesis. It targets queries that require cross-referencing multiple web sources before producing an answer.

- **Model ID:** `perplexity/sonar-pro`
- **Type:** chat
- **Providers:** perplexity
- **Context window:** 200,000
- **Maximum output tokens:** 8,000
- **Pricing:** See the provider details below
- **Canonical page:** https://vercel.com/ai-gateway/models/sonar-pro

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

```ts
import { streamText } from 'ai'

const result = streamText({
  model: 'perplexity/sonar-pro',
  prompt: 'Why is the sky blue?'
})
```

## About

Sonar Pro is the enhanced tier of Perplexity's search-augmented model family. It shares the same foundational architecture as Sonar, running a live web search on every inference call. The difference: it retrieves and processes significantly more sources per query. The result is longer answers backed by a broader evidence base.

Base Sonar handles straightforward factual lookups efficiently. Sonar Pro is built for queries that demand cross-referencing. Examples include regulatory changes across multiple jurisdictions, technical comparisons of competing frameworks, or market analysis pulling from diverse sources. The model evaluates more candidate sources, weighs conflicting information, and produces longer syntheses with richer citation coverage.

Sonar Pro keeps the same developer experience as base Sonar: a single API call returns a web-grounded, cited response with no external search configuration. The context window of 200K tokens supports extended multi-turn research sessions. If you already use base Sonar, upgrading to Pro is a model identifier change with no code modifications.

## What to consider

Sonar Pro searches more sources per request than base Sonar. This increases both response quality and latency. Profile response times under realistic query loads before setting user-facing timeout expectations. Web search calls bill separately (N/A per thousand when listed for this model).

## When to use

### Best For

- **Complex research queries** that require synthesizing information from many web sources into a single answer
- **Competitive analysis and market research** where breadth and depth of source coverage drive answer quality
- **Technical due diligence** comparing frameworks, products, or approaches across multiple authoritative sources
- **Regulatory and compliance research** that needs to cross-reference policies across jurisdictions or evolving standards
- **Knowledge work** where detailed, well-cited answers justify the higher per-token cost over base Sonar

### Consider Alternatives When

- The query is a simple factual lookup where base Sonar provides sufficient accuracy at lower cost
- You need explicit chain-of-thought reasoning over search results. Sonar Reasoning Pro adds structured deliberation
- Response latency must stay minimal. Base Sonar's lighter retrieval is faster for time-critical paths
- The task doesn't benefit from web grounding, such as pure code generation or creative writing

## Best for

- **Complex research queries:** Require synthesizing information from many web sources into a single answer
- **Competitive analysis and market research:** Breadth and depth of source coverage drive answer quality
- **Technical due diligence:** Compare frameworks, products, or approaches across multiple authoritative sources
- **Regulatory and compliance research:** Needs to cross-reference policies across jurisdictions or evolving standards
- **Knowledge work:** Detailed, well-cited answers justify the higher per-token cost over base Sonar

## Consider alternatives

- **Simple factual lookups:** Base Sonar provides sufficient accuracy at lower cost
- **Chain-of-thought needed:** Sonar Reasoning Pro adds structured deliberation over search results
- **Minimal latency required:** Base Sonar's lighter retrieval is faster for time-critical paths
- **No web grounding:** Pure code generation or creative writing doesn't benefit from search

## Frequently asked questions

### How does Sonar Pro differ from base Sonar?

Sonar Pro retrieves and processes more web sources per query, producing longer answers with broader citation coverage. Base Sonar is faster and cheaper for simpler factual queries.

### Does Sonar Pro include citations?

Yes. Like all Sonar models, responses include inline citations referencing the web sources consulted. Sonar Pro typically cites more sources per response due to its deeper retrieval.

### Is there an API difference between Sonar and Sonar Pro?

No. The API interface is identical. Switching from Sonar to Sonar Pro requires only changing the model identifier. You don't need code changes or configuration adjustments.

### What is the context window for Sonar Pro?

200K tokens. This supports extended multi-turn research sessions where context accumulates across exchanges.

### What does Sonar Pro cost?

Check the pricing panel on this page for today's numbers. AI Gateway tracks rates across every provider that serves Sonar Pro.

### How do I authenticate with Sonar Pro through AI Gateway?

Use your AI Gateway API key with the model identifier ``perplexity/sonar-pro``. AI Gateway handles provider routing and authentication automatically.

### When should I use Sonar Pro instead of a reasoning variant?

Use Sonar Pro when you need broad source coverage and detailed synthesis but don't need explicit step-by-step reasoning. For queries that benefit from chain-of-thought deliberation over search results, use Sonar Reasoning Pro.

## Links

- [Model page](https://vercel.com/ai-gateway/models/sonar-pro)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://sonar.perplexity.ai)
- [Provider pricing](https://docs.perplexity.ai/docs/pricing)


If you encounter incorrect information, can’t find what you need, need a capability Vercel doesn’t currently offer, or have any other feedback to report, please do it by following the instructions at https://vercel.com/feedback.md.