Perplexity Free vs Pro vs API 2026: What Does AI Research Actually Cost?
Perplexity's Free, Pro and Max plans buy use of its own research interface. Its API Platform is a separate pay-as-you-go product for software. The distinction matters more than a headline token price: a $20 Pro subscription does not give a developer a pool of included API calls, and an API key does not unlock Pro features in the web app. Perplexity's plan guide now explicitly says no complimentary API credits come with its API product.
What the personal plans buy
| Plan | Published US price | Practical reason to choose it |
|---|---|---|
| Standard (Free) | $0 | Basic cited search and occasional deeper work. |
| Pro | $20/month or $200/year | Frequent Pro Search, access to advanced model choices, more file analysis and limited Create files and apps use. |
| Max | $200/month or $2,000/year | Higher access for sustained Research, Create files and apps, advanced models and priority support. |
| Education Pro | $10/month with eligibility verification | Pro-style access for eligible students and educators. |
The Perplexity site publishes Pro at $20/month or $200/year; its Max help page publishes $200/month or $2,000/year, with annual Max billing on the web. Prices and app-store checkout terms can differ by location or platform, so check your actual payment screen before subscribing. Education Pro requires verification, according to the plan guide.
Free is not "no research": Perplexity's current plan matrix lists three Pro Searches per day and one Research query per month, alongside practically unlimited basic searches. Free has limited file uploads and does not offer the advanced-model selection available to paid plans. Pro expands Pro Search, Research, advanced model access and file work, but its consumer guide describes weekly or monthly limits based on average use, not a fixed promise of unlimited advanced queries. Max raises access further; it is not a sensible purchase solely to remove occasional Free friction. Exact available models and remaining allowances can change, so the account's current plan and usage screen is the final check.
Why Pro may be worth $20
Pro is buying a workflow: deeper cited answers, advanced model selection, larger file analysis, image and video generation, and some Create files and apps capacity in one interface. If you mainly ask quick factual questions, Free's basic search already covers that. If you regularly inspect sources, upload documents or need the model picker, the flat Pro fee may save time compared with juggling separate tools. A frequent heavy user should watch which specific quota runs out; a Research or Create limit can matter more than the count of ordinary searches.
Max is a tenfold jump in subscription price. Its official description focuses on higher-volume advanced research, creation, early access and support. There is no credible universal "queries needed to break even" against Pro because the plans give access to different limits and features, not a published per-query consumer tariff. Upgrade on observed usage, not a theoretical token calculation.
API pricing is a different bill
Perplexity's API rate card lists several different products. Choose one for the output your application actually needs:
| API path | Current billing shape | Use it when |
|---|---|---|
| Search API | $5 per 1,000 successful requests, or $1 with Fast Search; no token charge | You want ranked web results and will write your own answer logic. |
| Agent API | Model input/output tokens plus tools used | You want a model, grounded research or an agentic workflow. |
| Sonar Chat Completions | Sonar token rates plus a search-context request fee | You maintain an existing Sonar integration. |
For Search API, one successful POST /search containing up to five queries is one billed request. A successful request with zero results still counts; invalid, rate-limited and upstream-failed requests do not. One thousand standard requests therefore cost $5 whether they contain one query each or up to five each.
For Agent API, the listed perplexity/sonar model is priced at $0.25 per million input tokens and $2.50 per million output tokens. Standard web search adds $0.0025 per invocation, Fast Search $0.001 and URL fetching $0.0005. Tool calls are counted separately from model tokens. Third-party models are available too, but their rates and capabilities vary by model; the API model list says its prices may differ from provider list prices. Do not substitute a provider's direct rate for Perplexity's own Agent API price without checking the selected model row.
For the legacy Sonar Chat Completions product, the current rate card lists base Sonar at $1 input / $1 output per million tokens, plus a $5, $8 or $12 per 1,000 requests search-context fee for low, medium or high context. Sonar Pro and Reasoning Pro have different token and request rates. These fees are not the same as Agent API perplexity/sonar pricing.
Perplexity recommends migrating existing Sonar Chat Completions users to Agent API and using Agent API for new projects. Its current migration page also says Sonar Chat Completions remains supported. A migration changes request/response shape and possibly cost, so test both against your workload.
Cost per query, with the extras included
Assume 500 input tokens and 200 output tokens, no cache, one standard web-search invocation for Agent API, and low search context for legacy Sonar:
| Implementation | Token cost | Search/tool fee | Total |
|---|---|---|---|
Agent API perplexity/sonar + one web search | $0.000625 | $0.0025 | $0.003125 |
| Sonar Chat Completions, low context | $0.0007 | $0.005 | $0.0057 |
In the first row, 500 / 1,000,000 x $0.25 = $0.000125 input and 200 / 1,000,000 x $2.50 = $0.0005 output. The second row uses $1/$1 token rates and a $0.005 request fee. These are cost illustrations, not a claim that the two endpoints produce identical answers or make the same number of web searches. Multi-step Agent requests can invoke tools repeatedly; inspect each response's usage and cost data.
Pick by workflow, not by headline price
- A few cited answers each week: start with Free. Its basic search is useful before any payment.
- Frequent personal research and documents: test Pro at $20/month. Pay for its interface, model access and larger allowances, not for API credits.
- Sustained high-volume Research or creation: move to Max only after Pro's actual limits are the bottleneck.
- Search in your own product: use Search API if you need results, Agent API if you need generated, grounded work. Estimate model and tool calls together.
The related Perplexity Pro versus ChatGPT Plus guide covers the subscription decision against a general assistant; the DeepSeek cost guide covers a very different low-cost text API.
Frequently asked questions
Is Perplexity Free limited to three searches a day? No. The current plan guide says basic search is practically unlimited; Pro Search is listed at three per day on Free.
Does Pro include monthly API credits? No complimentary API credits are listed in the current plan comparison. API use is a separate pay-as-you-go purchase.
Is Sonar gone? No. Sonar Chat Completions is still supported in Perplexity's migration guide, and the Agent API also offers its own perplexity/sonar model. Check which interface and rate card your code actually uses.