Sonar
Sonar is Perplexity's in-house model, now offered as perplexity/sonar in the Agent API alongside third-party models. Perplexity ended support for the standalone Sonar Chat Completions API on September 27, 2026, and points those users to Agent API presets.
Specifications
- Context window
- —
- Max output
- Not published
- Input price
- $0.25per 1M tokens
- Output price
- $2.50per 1M tokens
- Cached input
- $0.062per 1M tokens
- Weights
- ClosedProprietary
- Parameters
- Undisclosed
- Input types
- Text
- Output types
- Text
- Reasoning mode
- No
- Tool calling
- Not documented
- Released
- Jan 2025Current
Agent API rates for perplexity/sonar; tool calls such as web_search are billed separately per 1,000 calls. The legacy Sonar Chat Completions API listed sonar at $1/$1 plus per-request search fees. Context window and max output are null because Perplexity does not document them for the Agent API version.
What it costs in practice
| Workload | Monthly cost at list price |
|---|---|
| Support chatbot30,000 conversations a month, 2,000 input and 400 output tokens each | $45 |
| Document summaries2,000 documents a month, 12,000 input and 800 output tokens each | $10 |
| Coding agent1,000 tasks a month, 60,000 input and 6,000 output tokens each | $30 |
Standard rates, no caching or batch discounts. Run your own numbers.
Strengths
- $0.25 input and $2.50 output per million tokens in the Agent API
- Cached input billed at $0.0625 per million tokens
- Built by Perplexity for lightweight, search-grounded question answering
Trade-offs to weigh
- The standalone Sonar API is legacy; support ended September 27, 2026
- Perplexity publishes few specs for this model, such as its context window
Where it fits
Perplexity positions Sonar for:
Where you can use it
- API
- OpenRouter
Alternatives to consider
| Model | Tier | Context | Input / output per 1M | |
|---|---|---|---|---|
| Gemini 3.5 Flash-LiteGoogle | Efficient | 1M | $0.30 / $2.50 | Compare |
| Gemini 3.1 Flash-LiteGoogle | Efficient | 1M | $0.25 / $1.50 | Compare |
| Grok Build 0.1xAI | Efficient | 250K | $1 / $2 | Compare |
| Codestral 25.08Mistral AI | Efficient | 125K | $0.30 / $0.90 | Compare |
Frequently asked questions
How much does Sonar cost?
Sonar lists at $0.25 per million input tokens and $2.50 per million output tokens, with cached input at $0.062 (checked October 10, 2026). Agent API rates for perplexity/sonar; tool calls such as web_search are billed separately per 1,000 calls. The legacy Sonar Chat Completions API listed sonar at $1/$1 plus per-request search fees. Context window and max output are null because Perplexity does not document them for the Agent API version. Confirm current rates on the provider's pricing page before you commit.
Is Sonar open source?
No. Sonar is a proprietary model from Perplexity; its weights aren't published. You can use it through the Perplexity API and through OpenRouter.
What is Sonar best for?
Perplexity positions Sonar for search, research, high volume. It accepts text as input, doesn't offer a separate reasoning mode and doesn't document tool calling. Test it on a sample of your own tasks before you choose it for production.
Sources
- docs.perplexity.ai/docs/agent-api/models
- docs.perplexity.ai/docs/getting-started/pricing
- docs.perplexity.ai/docs/getting-started/pricing.md
- docs.perplexity.ai/docs/agent-api/models.md
- docs.perplexity.ai/docs/agent-api/migrate-from-sonar/overview.md
- docs.perplexity.ai/docs/resources/changelog.md
- openrouter.ai/perplexity/sonar
Facts checked against these pages on October 10, 2026. Spot an error? Tell us.
Still choosing a model?
Send us your use case and constraints. We'll come back with a shortlist and an evaluation plan you can run on your own data.
- Free and vendor-neutral: no model provider pays for placement.
- Your details go only to partners you agree to.