Perplexity Pricing: Pro, Max, Teams and API Costs

Perplexity costs $0 to start, $20 a month for the plan most buyers need, and $200 a month at the top of the consumer ladder. Pro is $20 a month or $200 a year, Max is $200 a month or $2,000 a year 1 2, and the vendor's pricing card now prints only the annual-billing equivalents of those prices, $17 and $167 a month, with no monthly figure on the card 3. Workplace seats start at $40 a month or $400 a year per seat, with an Enterprise Max tier at $325 a seat per month 4 5. Developers pay on a separate meter: the Search API runs $5 per 1,000 requests or $1 per 1,000 on its lower-latency fast setting, Sonar model tokens run $1 to $15 per million 1, and a new Orchestrator API, renamed from Router, bills hosted open-weight models per token with no request fees 6. Verified students and educators can get every Pro feature for $10 a month 7.

Free gets you three Pro searches a day

The free tier costs $0. Its card line now reads "access to basic AI models," with "access to top AI models" printed one rung up on Pro 3; the plan guide's comparison table likewise marks "Advanced AI Models: No" for the free plan 8. The card's own FAQ still says Free "includes access to the top AI models with limited weekly usage" 3, so the vendor's surfaces currently disagree on where model access starts; the safe read is that Free guarantees the basic tier of models and nothing more. The caps are concrete as of the vendor's current plan guide, last modified 8 October 2026: 3 Pro Searches a day and 1 Research query a month 8; an independent tracker card verified in September prints the same 3-a-day Pro Search cap 9. That is enough to learn the product and enough to outgrow it. Anyone whose research runs daily hits the wall inside a week, and the cheapest fix is the $20 tier.

What $20 Pro adds, and what $200 Max adds on top

The ladder reads best as a delta rather than five feature lists. What money buys is search volume, Computer access and credits, model access, and organizational features; the free card's model line reads "access to basic AI models," Pro prints "access to top AI models," and Max prints "frontier AI models" 3.

• Pro, $20 a month or $200 a year 1 2, raises every limit and adds Perplexity Computer, with expanded access and a 4,000-credit bonus printed on the card 3, plus premium data sources like PitchBook; Pro Search moves from a daily cap to weekly limits, and Research queries move from 1 a month to monthly limits 3 8.
• Education Pro carries everything in Pro at $10 a month for verified students and educators, a 50 percent discount processed through SheerID 10 8 7.
• Max, $200 a month or $2,000 a year 1 2, adds maximum Computer usage, now 10,000 credits a month plus a 35,000-credit bonus on the current card 3, up from the one-time 20,000-credit bonus in the vendor's July 2026 release notes 11 9, expert-level research, frontier AI models, and the highest usage limits and memory 3; Mashable's coverage prints the same $200 and $2,000 figures 12.
• Enterprise Pro, $40 a month or $400 a year per seat, adds seat management and internal knowledge search through shared spaces 8; eligible educational institutions, nonprofits, and government agencies can claim 25 percent off, which lands a seat at $30 a month or $300 a year 4 8.
• Enterprise Max, $325 a seat per month, targets organizations that need advanced data security and collaboration tooling 4 5 2.

Every annual option is the same deal: two months free

Perplexity sells four annual prices and they share one shape. Pro falls from $240 paid monthly to $200; Max from $2,400 to $2,000; an Enterprise Pro seat from $480 to $400; and an Enterprise Max seat from $3,900 to the $3,250 now listed on the vendor's own enterprise pages 4 5 2. The pricing card itself no longer prints a monthly figure: it leads with the annual-billing equivalents, $17 a month for Pro and $167 for Max 3, which are the $200 and $2,000 annual prices spread over twelve months and rounded, as trackers printing the full prices confirm 1: $200 over 12 is $16.67, and $2,000 over 12 is $166.67. Each annual price is the monthly price times ten: two months free, a 16.7 percent discount, identical on every rung of the ladder. Perplexity's own billing FAQ prints the arithmetic for the Pro seat, calling it "a discount of approximately 16.7% versus monthly" billing 5. The enterprise onboarding guide, which previously labeled the same $40 and $400 numbers a "20% discount," now prints "17% discount" 4; both vendor pages now agree with the arithmetic, and whichever percentage a page prints, the price is the same ten months a year.

The API meters requests and tokens separately

Consumer plans bill seats; the API bills usage. The Search API charges $5.00 per 1,000 successful POST requests, or $1.00 per 1,000 on the new lower-latency fast search setting 13. One request may carry an array of up to five queries and still count as a single billing unit, invalid and rate-limited requests are not billed, and there are no token charges 13. Full batching drops the effective floor to $0.001 per query on standard search and $0.0002 on fast. The Sonar models, which generate answers rather than return links, bill twice:

• Token rates per million: Sonar $1 input / $1 output; Sonar Pro $3 / $15; Sonar Reasoning Pro $2 / $8; Sonar Deep Research $2 / $8 plus $2 for citation tokens, $3 for reasoning tokens, and $5 per 1,000 of the searches it runs on its own 1 9.
• Request fees per 1,000 requests scale with search context size: Sonar $5 low / $8 medium / $12 high, and Sonar Pro and Sonar Reasoning Pro $6 / $10 / $14; high context multiplies Sonar's request fee 2.4 times 1 9.
• The Agent API sells third-party models from OpenAI, Anthropic, Google, xAI, Z.AI, Moonshot AI, and NVIDIA at direct provider rates with no markup, plus tool charges: web_search $0.0025 an invocation standard or $0.001 with fast search, image search $0.0025, fetch_url $0.0005, people and finance search $0.005 each, and a sandbox at $0.03 per 20-minute session 13.
• The Orchestrator API, renamed from Router and still in private preview, sells Perplexity-hosted open-weight models per token with no request fees, with cache reads billed at each model's discounted rate 6 13; the hosted roster now lists DeepSeek, Moonshot AI, and Z.AI, from $0.15 input / $0.50 output per million on GLM-5.3 Flash to $3 / $15 on Kimi K3 6, and NVIDIA's models remain available through the Agent API at provider rates 13.
• Embeddings run $0.004 to $0.05 per million tokens by model 13.
• A Decisions API answers yes-or-no, multiple-choice, and scored questions at $0.01 to $0.02 per million input tokens by model, with output tokens free and no request fee 13.

The vendor's pricing page now leads with the Orchestrator and Agent APIs and publishes a migration path from Sonar to the Agent API; Sonar Chat Completions support ended on September 27, 2026, with synchronous and streaming requests kept working as they are reformulated into Agent API requests, and asynchronous Sonar requests no longer supported 14. The Sonar rates above have come off the vendor's pricing page with that retirement, and independent trackers, one updated 23 October 2026 and one verified in September 2026, print the same final card 1 9.

What one request actually costs once you do the math

The rate card prices parts, and buyers need totals. A Sonar Pro request with 1,000 tokens in and 1,000 tokens out costs $0.003 for input, $0.015 for output, and a $0.006 request fee at low context, about 2.4 cents in total. The same request at high context costs $0.032, and search depth alone raises a fixed-size request's price by a third without one extra token 1 9. For a grounded answer on the Agent API, the vendor's own worked examples land a web-search answer at $0.0059 and a low-effort research run at $0.0022 13.

That metering sets a ceiling on what a search is worth. One Pro month, $20, buys 4,000 standard Search API requests, 20,000 at the $1 fast tier, or up to 100,000 fast queries batched five to a request, but only about 830 generated Sonar Pro answers at the 2.4-cent floor. A grounded answer costs roughly five times a raw search, $0.024 against $0.005, and the $200 annual Pro plan converts to 40,000 standard searches or 200,000 fast ones at metered rates 1 2.

The API and the app are different products, so this is economics, not a shopping rule. A workload that needs fewer than 4,000 standard searches, or 20,000 fast ones, a month and no answer generation pays less metered than a $20 seat; anything heavier, or anything a person reads, belongs on the flat plans.

Which plan fits which reader

Free suits occasional lookups, with three Pro searches a day and one research query a month as the entire budget, the caps as of the vendor's current plan guide, last modified 8 October 2026. Pro is the default for anyone past daily use, and it is where Perplexity Computer access starts; students and educators should verify for the $10 rate before paying $20. Max is a credits product for buyers who burn through Computer usage and want frontier models and expert research on tap. Enterprise Pro seats fit teams that need shared knowledge search and administration, and Enterprise Max seats add the security and collaboration layer on top. The API is for building Perplexity into your own product, where raw search is cheap and generated answers are the cost center. Model access is now part of the dividing line, at least on the vendor's current pages: the card reserves "access to top AI models" for paid plans and lists "access to basic AI models" on Free 3, though the card's own FAQ still describes Free as including "the top AI models with limited weekly usage" 3. Search volume, Computer credits, and organizational features are the rest of the line. The cheapest rung that covers the limit you actually hit is the whole decision.

Verified October 2026 against Perplexity's pricing hub, its help-center billing articles, its API documentation, and current independent trackers, US billing; the vendor's consumer pages read cleanly this cycle, so consumer figures are pinned to the vendor's own pages where they print and to trackers where the card prints only annual equivalents. Change note, October 2026 refresh: no list price moved; the pricing card dropped its monthly figures for $17 and $167 annual-billing equivalents, Max's Computer bonus credits rose from 20,000 to 35,000, Pro's card line now carries a 4,000-credit bonus, the free tier's card line now reads "access to basic AI models," the hosted Orchestrator roster now lists DeepSeek, Moonshot AI, and Z.AI, and the enterprise onboarding guide changed its discount label from 20 percent to 17. This page is maintained: when Perplexity reprices, the figures above and a dated change note update here. To test whether a higher tier pays for itself against your own usage, run the same break-even math with your numbers in ProvenBrief's AI subscription calculator 15.