Perplexity Free, Pro or API: pay for the research workflow you need
Here's the quick version: pick Perplexity's app plans when you're the one doing the research, and its API products when your software is making the requests. Standard standalone Search costs $5 per 1,000 successful requests, so 10,000 requests cost $50. That usage is a separate purchase from the $20 monthly Pro subscription, so don't expect one to cover the other.
Start with the research task
If you only have the occasional question, start with the free Standard plan and see whether its search and file limits cover what you need. If they do, there's really no reason to add a subscription. Pro expands what you can do in the app, but an automated research pipeline needs the separately billed API instead.
The current selected Pro monthly payment in our price reference is $20. Choose the plan from the checkout available to your account, including tax and billing term. An API balance is its own purchase. The subscription comparison explicitly separates API access from web/app features, and it says there are no complimentary API credits on the API Platform route.
Search API: successful requests are the unit
| Standalone Search API route | 10,000 successful requests |
|---|---|
| Standard at $5 per 1,000 | $50 |
| Fast Search at $1 per 1,000 | $10 |
What to keep in mind: the documented billing unit is a successful POST /search request, not each query inside an allowed query array. A successful response can be billed even when it comes back with no results. So when you reconcile usage, check errors and request status rather than only counting the searches that gave you useful links. Your invoice and your "good results" tally won't always match.
Agent research has more than one meter
Perplexity's Agent API separates model-token charges from tool calls. Its listed standard web_search tool is $0.0025 per invocation and fetch_url is $0.0005. The practical catch is that these rates describe Agent tools, not the standalone Search API price above, so it's easy to mix them up.
Here's a scenario with 10,000 research jobs, where each job makes two standard web searches and three fetches. The tools cost 20,000 × $0.0025 + 30,000 × $0.0005 = $65. Then add the chosen model's input and output charges on top. And if the agent makes extra searches to resolve a tricky question, counting completed user jobs alone will undercount the bill.
Decide whether Pro replaces or duplicates another assistant
If ChatGPT or another tool already handles your research within an existing allowance, test whether Perplexity's source workflow saves enough review time to justify a second subscription. On the flip side, if finding sources is your real bottleneck, a search-oriented product can earn its place even when another assistant writes the nicer draft.
DeepSeek can be a metered model choice for an application, but a plain text-model call isn't the same thing as a tool-using search workflow. Include the search/fetch route and its charges before you compare total cost. The lowest token rate doesn't automatically buy you a useful source list.
A practical ten-question pilot
Pick ten questions your work actually raises. Mix in a recent factual question, a niche product detail, a comparison and something that needs a primary document. For each one, note the original source behind every important claim, whether the cited page really says it, and what corrections you needed before you could use the answer.
Then compare accepted research outputs, elapsed review time and full payment. Open the cited page and check that it supports the actual claim. A citation is a lead to inspect, not proof that an answer is correct. Keep the questions and dated results, so your next purchasing decision rests on your own workflow rather than a generic leaderboard.
The worksheet below prices standalone Search requests. Switch its rate to $0.001 for the documented Fast Search route. It doesn't include Agent model tokens or tool calls, so those need the separate calculation above. For application pricing, carry the complete job cost into the unit-economics model.
Price the standalone Search API workload
Starts with standard Search at $5 per 1,000 successful requests. Enter $0.001 per request for the documented Fast Search route. This subtotal excludes Agent API tokens and tools and does not assume Pro supplies API credits.
Plan and rate reference
| Provider | Plan or rate | Price | Billing |
|---|---|---|---|
| Perplexity | Pro | $20 | account / month · Monthly |
| Perplexity | Enterprise Pro — starting rate | $40 | seat / month · Monthly |
Compare research-tool subscriptions and API routes →
Frequently asked questions
Is Perplexity Pro worth $20 per month compared to the Free tier?
Perplexity Pro is worth the $20/month subscription if you perform more than 5 complex research queries per day, need access to top reasoning models like Claude 3.5 Sonnet or OpenAI o1, or require document upload analysis. For occasional search replacement, the Free tier's unlimited Quick searches are sufficient.
How does Perplexity Sonar API pricing compare to direct OpenAI or Anthropic APIs?
Perplexity Sonar API charges a combined fee: per-token rates (ranging from $1 to $5 per million tokens depending on model size) plus an online citation query surcharge (typically $5 per 1,000 online searches). If you already implement your own retrieval pipeline, direct model APIs are cheaper; if you need turnkey web grounding, Sonar saves substantial engineering hours.
Does Perplexity Pro include free API credits?
Perplexity Pro subscriptions historically included $5 in monthly API credits on Sonar endpoints for developer experimentation, but heavy production deployments require billing payment methods linked to the Perplexity API platform.