Check the API project’s actual credit and any explicit offer before treating Perplexity requests as free. Keep consumer subscriptions, developer billing and usage tiers as separate conditions.

What is free at Perplexity right now

The official billing documentation was checked. No universal ongoing free inference allowance was established by that source. Promotional or account-specific credits require a check in your account. Read the official terms ↗

Last verified · Source ↗

The public API quickstart describes pay-as-you-go access, and the project guide describes credit-based billing. Those sources do not establish a universal ongoing free inference allowance. Verify any promotion directly in the intended API project before relying on it. Official Perplexity documentation. Official Perplexity documentation.

Be precise about the entitlement you are checking. A consumer subscription, a temporary API credit, a usable purchased balance and an account’s admission tier answer different questions. An old article about a subscription benefit is not enough evidence that a newly created key has usable credit today. Keep the offer’s applicable product and terms in your own test record.

Do not confuse a public documentation example with a free execution environment. The sample request explains the API contract; running it uses the account configured by the key. Decide who owns the experiment and how much work it will perform before connecting a public input form or an unattended research loop.

Limits of the free tier

Published limits have not yet been verified for this selection. Check the official limits documentation and your account console.

Last verified · Source ↗

An empty free-plan table does not establish unlimited use. Perplexity publishes usage tiers and product-specific admission rules, which you can inspect in the rate-limit guide. A tier can describe what requests an account may submit without proving that its current balance funds those requests.

If your account has an explicit offer, check which endpoint and model it covers. A benefit tied to one product should not be generalized to Router, Agent, Search and Sonar alike. Preserve any applicable restrictions rather than presenting a single free-requests figure that hides the product distinction.

For research work, keep a bounded list of questions and a clear stopping condition. A follow-up question can trigger additional model or search work even when the user sees one continuous conversation. Record completed tasks and returned usage so you can explain what the experiment consumed before extending it.

What ends the free tier

Use the actual offer’s terms to determine expiration, exhaustion and eligible usage. This page does not invent a trial duration or a subscription credit amount. If the terms cannot be verified, treat the allowance as uncertain and plan the experiment under the published paid structure.

Inspect automatic reload in the project’s billing preferences. The official guide describes automatic credit funding as an available setting. If you want a trial to stop before purchasing more credit, understand that setting before dispatching traffic; do not assume the application will naturally stop at the boundary you had in mind. Official Perplexity documentation.

Define an administrative outcome for a funding problem. Stop new dispatches, retain safe unfinished tasks and explain whether the user’s request was completed or deferred. Avoid a generic retry loop that keeps treating an account condition as a temporary network failure. Review Perplexity error handling for the separate failure categories.

The Sonar support deadline is also relevant to an evaluation plan. The official Sonar documentation states that Sonar support continues until September 27, 2026 and directs developers to the Agent API migration guide. Official Perplexity documentation. A trial that proves a near-retirement interface works is not sufficient preparation for a new long-lived application. Include the current Agent contract in the evaluation if that is the intended production path.

How to build within the free tier

Use local fixtures for application behavior that does not require live generation. Test input validation, credential-presence checks, output rendering and safe error presentation without paying for every UI iteration. Label example responses clearly so screenshots and test logs cannot be mistaken for evidence of a successful account call.

Choose evaluation questions that teach you something useful. A documentation question with an identifiable primary source can reveal whether search and citation handling fit the application. Include an ambiguous question and one whose requested fact is absent. Decide in advance what a useful answer should do in each case, then record accepted and rejected outcomes rather than only successful connections.

Use the AI API cost calculator for supported model-token assumptions, and keep additional product charges outside the estimate until you can model them explicitly. A token-only figure should not be treated as the total cost of an Agent research run. Inspect returned usage and the account record after a representative task before increasing volume.

Follow the product’s prompting and search controls. For Sonar, the documented filter parameters belong in the request contract; a prompt that mentions a restriction is not a substitute for configuring the supported filter. Test that your chosen restrictions still allow the evidence needed to answer the question. Official Perplexity documentation. Official Perplexity documentation.

Keep the final artifact focused. If the application needs a short answer with source references, ask for that structure and assess whether the evidence supports it. More generated explanation is not automatically a better trial result. Preserve the sources and completion state alongside the answer so a reviewer can inspect why the application accepted the task.

Set up the account through Perplexity API key instructions and use the Python walkthrough for a controlled current-product call. Review Perplexity charge categories when changing products or configurations. A small evaluation should leave you with a reproducible integration and an understandable consumption record.

Review the trial as a complete workflow rather than a collection of attractive answers. Record which questions required clarification, which sources were missing and which results were rejected during evidence review. Include the work of finding and correcting those outcomes when deciding whether to continue. This makes the evaluation useful even if an explicit credit offer ends before every planned task is complete. The retained fixtures and decisions can still guide a deliberate paid run or a comparison with another provider.

Alternatives if you need more

Compare OpenRouter zero-priced model routes and Google’s selected-model free tier using their own official eligibility and limits. If search-backed answers are essential, check whether the substitute actually supplies the required evidence workflow rather than only a text-generation model.

Use the same safe task and acceptance criteria across candidates. A free model that cannot retrieve the needed sources may require an additional search service, which changes both the implementation and the cost. Keep those extra steps explicit. Return to the Perplexity API overview if the integrated product remains the best fit after that comparison.

Frequently asked questions

Does a consumer Perplexity subscription prove API credit?
No. Verify the actual API project and current offer terms. Consumer and developer access should be checked separately.
Is a new usage tier the same as free requests?
No. Admission tier and usable funding are different account conditions. Inspect both before dispatching work.
Can automatic reload turn an evaluation into paid traffic?
Review the project’s actual preference. Funding behavior and your application’s task allowance should be deliberate settings.
Does the free-tier table cover every Perplexity product?
Only verified applicable facts are shown. Do not generalize an offer or a missing row across Router, Agent, Search and Sonar.
What is a useful low-cost evaluation question?
Choose a safe representative task with checkable evidence and include an absent-information case. Measure accepted output, not only connectivity.
Should a new free trial use the legacy Sonar endpoint?
Review the announced support deadline and current Agent migration path. Evaluate the interface intended for your production application.

Sources

Last verified · Source ↗