Perplexity AI pricing

About

Perplexity AI is a San Francisco-based AI company that operates an answer engine — a hybrid between a search engine and an AI chatbot that surfaces real-time, cited answers rather than a list of links. Founded in 2022 by Aravind Srinivas (ex-OpenAI), Denis Yarats, Johnny Ho, and Andy Konwinski, Perplexity positions itself as the successor to the traditional search experience: ask a question, get a synthesized, source-cited answer powered by multiple frontier AI models.

By early 2026 Perplexity reports approximately $200M in annual recurring revenue and a $21B valuation following its Series E-6 funding round. The company has raised over $1.5B total from investors including Nvidia, Jeff Bezos, SoftBank, and IVP. It processes hundreds of millions of queries per month across web, iOS, and Android, with particular traction among researchers, students, and knowledge workers.

Perplexity competes directly with Google Search, ChatGPT, Claude, and Microsoft Copilot, while differentiating through transparent citations on every answer, real-time web grounding, and a model-picker that gives subscribers access to every major frontier model (OpenAI, Anthropic, Google) inside one $20/month subscription — a value proposition no single-vendor AI assistant can match.

Pricing summary : How Perplexity AI’s freemium + API stack works

Perplexity runs a freemium subscription model layered on top of a separate usage-based developer API. Consumer tiers are flat-rate monthly or annual subscriptions — Free, Pro ($20/mo), and Max ($200/mo) — gated by query quality (Pro Search quota) and model access rather than raw query volume. Enterprise tiers are seat-based, priced (as of the 2026-08-14 redesigned enterprise page) at $34–$271/seat/mo when billed annually, plus a custom-priced “Additional security and governance” add-on layer; governance features (SCIM, audit logs, RBAC, data retention) are now marked “Contract required” for both Enterprise tiers rather than gated by the previous 50-seat threshold.

The developer-facing API platform is entirely separate: pure usage billing, no seat or subscription minimum, split across five metered surfaces, one of which (Sonar) is now sunsetting. The Sonar API bills tokens plus a per-request search fee that scales with search-context size ($5–$14 per 1,000 requests); as of 2026-08-14 its docs state Sonar “will be supported until September 27, 2026” before full migration to the Agent API. The Search API bills requests only — $5.00 per 1,000, with no token cost at all. The Agent API resells third-party frontier models “at direct provider rates with no markup” and meters tools separately ($0.0025 per web_search, $0.03 per sandbox container session, and fetch_url at $0.0005 — this reverted to its original level by 2026-08-14 after a 2026-07-29 halving; see Pricing evolution). The Embeddings API bills per million tokens ($0.004–$0.05 depending on model). Launched 2026-08-11 as the “Gateway API,” this fifth surface gave “unified access to open-weight models hosted by Perplexity through a single endpoint and a single API key” — a catalog that grew from 3 models (DeepSeek, Kimi, GLM) to 5 by 2026-08-14 with two NVIDIA Nemotron additions, priced by Perplexity itself from $0.0115/1M input tokens, structurally distinct from the Agent API’s at-cost resale. By 2026-08-25 the docs had renamed it the “Router API” and gated it behind a “private preview” banner requiring an email request for access — the model catalog and every price are unchanged, but it is no longer self-serve. This dual-stack design — flat subscriptions for end-users, pay-per-use for developers — mirrors the freemium + API usage model increasingly common across AI platform companies.

What makes this different: The multi-model picker inside a single $20 subscription effectively collapses three separate AI tool subscriptions (OpenAI, Anthropic, Google) into one, making the pricing anchor unusually generous relative to alternatives. The Max tier mirrors ChatGPT Pro’s $200/month price point exactly — a deliberate competitive signal, not a coincidence.

Pricing by product

Consumer subscriptions (Individual plans)

Computer credits are the consumer-side usage meter: Perplexity states “100 credits equals $1”, which puts a hard dollar value on each plan’s monthly allowance. Credits are separate from API credits — “a Perplexity subscription is not required to purchase API credits or use the API,” and API credits cannot be spent on Computer.

Consumer subscriptions (Education / Nonprofit)

Enterprise subscriptions (Business plans)

perplexity.ai/enterprise/pricing came back online on 2026-08-14 (200, first-party) after sitting behind a Cloudflare challenge since at least 2026-07-21, with a redesigned page. It shows only a per-seat-per-month-when-billed-annually price for each tier — no separate monthly-only rate or billing toggle is visible — and drops the previous absolute weekly/monthly Pro Search, Research query, and Browser Agent query quotas in favor of multipliers over the base Pro plan (e.g., file uploads 30x Enterprise Pro / 100x Enterprise Max; asset generation and Comet Agent 2x / 20x).

Sonar API — token pricing (Developer / Platform)

Sunset notice (added 2026-08-14): Perplexity’s docs now state “Sonar Chat Completions is now Agent API. Sonar will be supported until September 27, 2026. Migration guide here.” The migration banner existed before (seen 2026-07-29 and 2026-08-11) but did not previously carry an end-of-support date. No rate below has changed as a result of this notice.

Sonar API — request fee by search-context size (per 1,000 requests)

Total cost per query = token costs + request fee. The request fee applies to Sonar, Sonar Pro, and Sonar Reasoning Pro only; Sonar Deep Research instead meters citation tokens, reasoning tokens, and search queries. Pro Search (a Sonar Pro mode requiring stream: true) carries its own request fee by search type: fast (default) at $6 / $10 / $14 per 1K, pro (multi-step tool usage) at $14 / $18 / $22 per 1K, and auto “varies by classification” — token pricing stays at the standard Sonar Pro $3 input / $15 output per 1M.

Search API (Developer / Platform)

Agent API — tool pricing (Developer / Platform)

Router API, formerly “Gateway API” — Perplexity-hosted models (Developer / Platform)

Access change (confirmed 2026-08-25): The docs now show a banner reading “Router API is in private preview. Email [email protected] to request access.” The product was called “Gateway API” through at least 2026-08-14; by 2026-08-25 every reference in the docs sidebar, page titles, and code samples read “Router API” instead, and the page is no longer presented as self-serve. The URL (docs.perplexity.ai/docs/gateway/models) and every model price below are unchanged.

Launched 2026-08-11 with 3 models; the catalog expanded to 5 by 2026-08-14 with the addition of two NVIDIA Nemotron models. The Router API (formerly Gateway API) is a distinct product from the Agent API: it exposes open-weight models “hosted by Perplexity” (the perplexity/ prefix marks who serves the model, not who created it) rather than reselling third-party frontier models at cost. Every model is reachable via both an OpenAI-compatible Chat Completions endpoint and an Anthropic-compatible Messages endpoint under the same creator/model-name id, billed “at the requested model’s rates regardless of how the request is served.” Cached input reads bill at the discounted cache-read rate shown; cache writes bill at that rate too (or at the input rate where no dedicated write rate is listed), and reasoning tokens bill at the output rate. No per-request fee. The 5 models’ prices are unchanged since 2026-08-14; the only change confirmed by 2026-08-25 is the rename and the private-preview access gate.

Embeddings API (Developer / Platform)

Sales motions across products: PLG / self-serve for Free, Pro, Max, and four developer APIs (Search, Agent, Sonar, Embeddings); request-access (private preview, email [email protected]) for the Router API as of 2026-08-25; sales-led for Enterprise Pro, Enterprise Max, and API credits purchased through AWS Marketplace or the enterprise inquiry form.

Hidden costs : What Perplexity AI users actually pay beyond the base plan

Archetype A: Solo researcher on Pro plan

A solo knowledge worker using Pro for daily research, file uploads, and occasional API testing:

Sonar API spend now sits fully outside the subscription: Perplexity’s current help-center billing article describes API credits as purchase-only (“API credits are available for purchase on a pay-as-you-go basis”), separate from the Computer credits that power the consumer product, and notes that “a Perplexity subscription is not required to purchase API credits or use the API.” Production API usage compounds fast on Sonar Pro, where output at $15/1M tokens is 15× more expensive than input.

Archetype B: 100-person Enterprise Pro team

A mid-size company on Enterprise Pro (annual billing), building Perplexity into internal knowledge workflows:

Note: SCIM provisioning is no longer gated by seat count. As of the 2026-08-14 enterprise page redesign, SCIM, audit logs, RBAC, and data retention are marked “Contract required” on both Enterprise Pro and Enterprise Max ($271/seat/mo, billed annually) — replacing the previous 50-seat threshold. Every team, regardless of size, now needs a signed contract to turn these on, which is itself a hidden cost (sales-cycle time) this dollar total doesn’t capture.

Want to estimate your own Perplexity AI bill? Use the Perplexity AI pricing calculator to model your monthly cost based on seat count and API usage volume.

Pricing evolution : Perplexity AI’s pricing history from free search to $200/mo Max

Cadence

Tracked range: 2022 Q3–2026 Q3. Quarters not listed above were verified stable (0 price changes, 0 SKU additions).

Notable changes

• 2024-01-01 — Pro plan launched at $20/month; free tier capped at ~5 Pro Searches/day; first revenue-generating product.
• 2024-07-30 — Publisher revenue-share program announced after Forbes and Wired plagiarism accusations; partners receive a percentage of ad revenue when their content is cited. CNBC coverage.
• 2025-01-22 — Sonar API launched as replacement for pplx-api, adding real-time search integration and citation support to the developer offering.
• 2025-07-02 — Max plan launched at $200/month, matching ChatGPT Pro’s price point exactly. TechCrunch coverage.
• 2026-01-01 — Developer pricing split in two: a standalone Search API at $5.00 per 1,000 requests with no token cost, and an Agentic Research API reselling OpenAI, Anthropic, Google, and xAI models “at direct provider rates with no markup.”
• 2026-01-15 — Enterprise page restructured into a Free / Enterprise Pro / Enterprise Max matrix, formalizing the higher-compliance Enterprise Max tier above the $40/seat Enterprise Pro plan.
• 2026-03-18 — Comet browser paywall dropped; Comet Plus add-on ($5/mo) introduced for publisher content.
• 2026-07-10 — perplexity.ai/pricing began returning a hard 404 (first archived on this date, still 404 when re-checked on 2026-07-21), taking consumer and seat prices off any public marketing surface.
• 2026-07-21 — Agent API gained a sandbox tool at $0.03 per container session — a 20-minute billing window rather than a runtime cap, with SDK searches inside it billed at $0.005 each — and widened zero-markup model resale to add Z.AI, Moonshot AI, and NVIDIA. No published rate moved: Sonar tokens, request fees by search-context size, the $5.00-per-1,000 Search API, and Embeddings all held.
• 2026-07-29 — Agent API web_search and fetch_url tool prices were both cut in half — to $0.0025 and $0.00025 per invocation — with the identical cut applied to SDK search queries run inside the sandbox tool (now $0.0025 per request, “same as web_search”). No other Agent API meter moved and no Sonar, Search API, or Embeddings rate changed; the docs’ own worked cost examples now total to the lower figures, corroborating the cut across three independent page renders.
• 2026-08-11 — Perplexity’s developer docs added a fifth API surface, the Gateway API: “unified access to open-weight models hosted by Perplexity through a single endpoint and a single API key,” exposed via OpenAI-compatible Chat Completions and Anthropic-compatible Messages. The initial catalog is three Perplexity-hosted open-weight models — perplexity/deepseek-v4-flash-0731 ($0.13/$0.26/$0.028 per 1M input/output/cache-read tokens), perplexity/kimi-k3 ($3.00/$15.00/$0.30), and perplexity/glm-5.2 ($1.40/$4.40/$0.14) — billed per token with no per-request fees. This is structurally different from the Agent API, which resells third-party models “at direct provider rates with no markup”: Gateway models are hosted and priced by Perplexity itself, so this is the first Perplexity developer product with real hosting margin built into the rate card. No Sonar, Search API, Agent API tool, or Embeddings rate moved.
• 2026-08-14 — Agent API fetch_url tool price reverted to $0.0005 per invocation, undoing the halving to $0.00025 that took effect on 2026-07-29 — confirmed by the docs’ own worked “Agent API Research Preset” cost example, which rose from $0.00675 to $0.007. web_search ($0.0025) and every other Agent API meter held.
• 2026-08-14 — perplexity.ai/enterprise/pricing came back online after sitting behind a Cloudflare challenge since at least 2026-07-21, relaunching with annual-only per-seat pricing ($34 Enterprise Pro, $271 Enterprise Max) and a restructured governance gate: SCIM, audit logs, RBAC, and data retention are now “Contract required” on both tiers, replacing the previous 50-seat threshold that applied to Enterprise Pro alone. A new custom-priced “Additional security and governance” card was added alongside the two named tiers.
• 2026-08-14 — The Sonar API’s migration banner gained an explicit end-of-support date: “Sonar will be supported until September 27, 2026.” No Sonar token, request-fee, Search API, or Embeddings rate changed.
• 2026-08-14 — The week-old Gateway API catalog grew from 3 to 5 Perplexity-hosted models, adding two NVIDIA Nemotron models — perplexity/nemotron-3.5-lightning-30b-a3b ($0.0115/1M input, now the cheapest model in the catalog) and perplexity/nemotron-3-ultra-550b-a55b ($0.25/1M input). The 3 original models’ prices were unchanged.
• 2026-08-25 — The Gateway API was renamed “Router API” throughout the docs sidebar, page titles, and code samples, and gained a banner reading “Router API is in private preview. Email [email protected] to request access.” The URL and all 5 model prices are unchanged from the 2026-08-14 catalog, but the product is no longer self-serve — the first Perplexity developer API to move from GA back to a gated preview rather than the reverse. Separately, perplexity.ai/enterprise/pricing — back online since 2026-08-14 — reverted to a Cloudflare challenge again, a capture-access regression rather than a confirmed price change; the $34/$271 seat rates are carried forward unverified.

The 2024 plagiarism controversy in detail

In June 2024 Forbes discovered a near-verbatim copy of a paywalled article surfaced in Perplexity’s Pages tool with no clear attribution beyond a small logo. Wired independently found an IP address “almost certainly linked to Perplexity” had visited its parent company’s sites over 800 times in three months, bypassing crawl disallowance. The backlash escalated: Dow Jones and The New York Post filed suit for copyright infringement in October 2024; News Corp renewed its case; CNN filed suit in May 2026 over alleged AI-powered content scraping.

Perplexity’s response was the July 2024 publisher revenue-share program — a direct commercial answer to the legal and reputational pressure. By Q1 2026, over 2,400 publishers participate, with a $42.5M payout pool under the Comet Plus rebrand. The incident illustrates how a citation-first product can face existential copyright risk when the citation pipeline resembles scraping rather than linking.

The disappearing consumer price page in detail

The direction of travel across 2026 is unusual: the developer rate card got more granular while the consumer one vanished, briefly reappeared, then vanished again. perplexity.ai/pricing has returned “Page not found. Error 404.” since at least 2026-07-10, and perplexity.ai/enterprise/pricing sat behind a Cloudflare bot challenge at every check from at least 2026-07-21 through 2026-08-11. perplexity.ai/pro loaded cleanly with no challenge on 2026-07-29 after being unreachable at every prior check, briefly restoring a first-party surface for Pro and Max pricing — but by 2026-08-11 it was back behind the same Cloudflare challenge, including under a stealth browser identity, and remained blocked as of 2026-08-14. By that same date, perplexity.ai/enterprise/pricing did the opposite: it came back online, unannounced, with a redesigned page. Three URLs, three independent outage/recovery timelines, none explained — this now reads as a pattern rather than a one-off incident.

The practical effect has narrowed but not closed. Free, Pro, Max, Enterprise Pro, and Enterprise Max prices are all unchanged throughout (within rounding — see below), and every one of them remains confirmable in first-party text even when a marketing surface is down: Perplexity’s help center states the Max rate (“$200/monthly”, “$2000/annually”), the Enterprise rates as last quoted on 2026-07-29 (“$40 per month or $400 per year”, “$325 per month or $3,250 per year”), and the Education Pro rate (“$10/month (50% off)”); Apple’s App Store listing independently shows the Pro in-app purchases at $20.00 and $200.00. As of 2026-08-14, one of the three original consumer/enterprise marketing URLs renders for an unauthenticated visitor — /enterprise/pricing, back online with its redesign — while /pricing and /pro remain blocked. For a company whose product promise is “every answer carries a verifiable citation,” a pricing surface that flips between broken and working without notice is itself the finding: procurement teams evaluating Enterprise once again have a first-party page to check against, but individual buyers pricing out Pro or Max still do not.

The redesign that came back is not a like-for-like restoration, either. The page now shows only an annual-billed per-seat rate ($34 Enterprise Pro, $271 Enterprise Max) with no monthly-only price or toggle, and it replaces the old 50-seat (or one Enterprise Max seat) threshold on SCIM/audit logs/RBAC/data retention with a blanket “Contract required” label on both tiers — trading a volume-based self-serve gate for a sales-gated one. A prospect can once again see the number, but can no longer self-serve their way to governance features at any seat count; every compliance-sensitive buyer now needs a signed contract regardless of team size. See What’s unique and Areas to improve for the buyer-impact read.

By 2026-08-25 even that recovery was short-lived: perplexity.ai/enterprise/pricing reverted to the same Cloudflare challenge, unannounced, just 11 days after coming back online. The $34 Enterprise Pro and $271 Enterprise Max seat rates in this article are now carried forward from the 2026-08-14 check rather than independently confirmable — the same evidentiary posture Free, Pro, and Max prices have been in since July. Four URL-level incidents on this domain in under two months, split evenly between outages and unexplained recoveries, is no longer plausibly a one-off scraping-defense misconfiguration; it reads as a pricing surface Perplexity is not treating as public infrastructure that has to stay up.

What’s unique : Perplexity AI’s distinctive pricing mechanics

1. Multi-model picker at a single flat rate. Perplexity Pro bundles access to GPT-5, Claude Opus 4, and Gemini 3.1 Pro inside one $20/month plan — a model-selection utility that would cost $60–$80/month if purchased from each vendor separately. This makes Perplexity the only AI subscription where the pricing arbitrage is the product itself. See how AI companies structure multi-model access for context on why this is unusual.

2. Three metering axes on one developer platform — tokens, requests, and now time. Unlike OpenAI or Anthropic, which charge per token only, Sonar bills both per token and per search request, the fee ($5–$14 per 1,000, scaling with search-context depth) reflecting the real-time retrieval cost that makes Sonar answers current. On 2026-07-21 Perplexity added a third axis that neither of the first two can express: the Agent API’s sandbox tool bills $0.03 per container session, where a session is explicitly “a 20-minute billing window, not a runtime cap.” The reason the meters keep multiplying sits on the same rate card — Perplexity resells third-party models from seven providers (OpenAI, Anthropic, Google, xAI, Z.AI, Moonshot AI, NVIDIA) “at direct provider rates with no markup,” so every dollar of margin has to come from the retrieval and execution wrapped around those tokens rather than the tokens themselves. A fourth wrinkle arrived on 2026-08-11: the Gateway API sells access to Perplexity-hosted open-weight models (DeepSeek, Kimi, GLM, and — as of 2026-08-14 — two NVIDIA Nemotron additions) at rates Perplexity itself sets, not at-cost resale. It’s the first Perplexity developer product with real hosting margin built into the card, sitting structurally apart from the Agent API’s zero-markup model tokens. That fourth axis went private almost as fast as it appeared: by 2026-08-25 the product was renamed “Router API” and gated behind a “private preview” banner requiring an email request, ending its two-week run as a self-serve surface. None of the five model prices moved — only who can act on them did. For a developer, that means cost-modelling multiple unrelated units and pricing philosophies at once (assuming access is granted at all), and the usage-based billing complexity is correspondingly higher than single-dimension APIs.

3. Compliance features gated behind a sales contract, not a seat count. Until 2026-08-14, Enterprise Pro SCIM provisioning unlocked only at 50 seats — a capability gate tied to volume, not tier. The redesigned enterprise page replaces that with a blanket “Contract required” label on SCIM, audit logs, RBAC, and data retention for both Enterprise Pro ($34/seat/mo) and Enterprise Max ($271/seat/mo, billed annually), so the mechanic shifted from a volume threshold to a universal sales gate: no team, of any size, can self-serve its way to governance features anymore. That’s arguably fairer than a seat cliff but strictly less self-serve, and it’s still unusual in enterprise SaaS pricing relative to, say, Intercom or Salesforce, where compliance add-ons are typically priced and purchasable without a bespoke contract.

4. Publisher revenue-share as a pricing-adjacent commercial model. The Comet Plus add-on ($5/mo) and the underlying publisher program create a revenue layer that does not exist in any other AI subscription product. Subscribers pay Perplexity; Perplexity shares 80% of that $5 with publishers whose content was cited. This is effectively a micro-micropayment content marketplace dressed as a subscription add-on — a novel monetization mechanic with significant legal and structural uncertainty.

5. Free tier as the primary growth engine. Perplexity’s Free plan is genuinely useful — unlimited basic searches, automatic source citations — making it one of the most competitive free AI tiers in the market. The 3-Pro-Search-per-day cap (tightened from 5/day in 2024) is a soft conversion nudge rather than a hard paywall. This PLG-first approach drove Perplexity from zero to $200M ARR with minimal sales investment.

Strengths & weaknesses

Billing UX : Perplexity AI’s subscription controls and payment experience

• Self-serve upgrade / downgrade — Plan changes are handled entirely in-app at perplexity.ai/settings/subscription. No sales call required for Pro or Max.
• Annual vs monthly toggle — Offered at the point of subscription; annual prices and the savings percentage are shown side-by-side at selection. Since perplexity.ai/pricing began returning a 404, the toggle is only reachable inside the product, not from a public marketing page.
• Auto-renewal — All subscriptions auto-renew. There is no pause option; users must cancel and resubscribe.
• Refund window — Non-EU users: 24-hour window for monthly plans; 72-hour window for annual plans. EU, UK, and Turkey users: 14-day statutory refund period for all plans.
• Payment methods — Credit card, PayPal, Google Pay, Apple Pay. Enterprise plans add wire transfer and invoice billing.
• Sonar API billing — “API credits are available for purchase on a pay-as-you-go basis,” billed to a prepaid balance with an optional Automatic Top Up that fires when the balance falls below $2. Perplexity states “a Perplexity subscription is not required to purchase API credits or use the API,” and its current help-center plan comparison lists no bundled monthly API credit on Pro (the $5/mo Sonar credit Pro carried in 2024 is no longer documented).
• Computer credits (consumer usage meter) — Separate currency from API credits, priced at “100 credits equals $1”. Monthly allocations: Max 10,000, Enterprise Max 15,000, Enterprise Pro 500; consumer Pro has no monthly allocation but receives 4,000 bonus credits at signup. As of 2026-07-29, perplexity.ai/pro also advertises a “LIMITED TIME” 35,000-credit signup bonus on Max (on top of its 10,000/mo allocation), badged as ”+$450 free Computer credits” — not present at the prior 2026-07-21 check. Top-up credit packs are sold as in-app purchases on iOS at the same 100-credits-per-dollar rate.
• “Estimate your cost” calculator — The developer pricing docs now embed a live cost estimator with per-surface tabs (Search API, Agent API, Sonar API, Embeddings API), a volume slider, and a “Show breakdown” expander. It is explicitly framed as non-authoritative: “Estimate only. Final cost is metered from each response’s usage field - view it in the API Portal.”
• Per-response usage metering — Every API response carries a usage field that is the authoritative billing record, viewable in the API Portal.
• API Groups & Billing — Admin console section for grouping API keys and attributing spend, alongside API Key Management and a Computer Analytics API for agent-level usage reporting.
• Rate Limits & Usage Tiers — Separate docs page governing throughput; spend controls and rate limits are managed as distinct surfaces.
• SheerID student verification — Education Pro ($10/mo) requires identity verification through SheerID; no self-attested discounts.
• Enterprise seat management — Admin portal for seat provisioning, role-based access, and SSO configuration. As of the 2026-08-14 redesigned enterprise page, SCIM, audit logs, role-based access control, data retention configurability, and a separate enterprise domain are marked ”✓ Contract required” for both Enterprise Pro and Enterprise Max — included in the plan but requiring a signed contract to activate, replacing the prior language that gated these behind a 50-seat (or 1 Enterprise Max user) threshold. A separate “Additional security and governance” custom-priced card bundles the same contract-gated features plus volume discounts and forward-deployed support.
• Procurement paths — API credits can be purchased through the AWS Marketplace listing (“consolidated billing and enterprise procurement”) or via the enterprise inquiry form for custom pricing.
• Spend visibility — API usage dashboard in the developer console shows token consumption and request counts per model; no predictive spend alerts or configurable spend caps documented.
• Router API (formerly Gateway API) model catalog / allowlist, now request-gated — GET /models at https://api.perplexity.ai/router/v1/models lists every available Router model with its base token prices; requesting a model outside the catalog returns a 400 naming the invalid model, so “the catalog is also the allowlist.” One API key covers both the OpenAI-compatible /chat/completions and Anthropic-compatible /messages endpoints under the same creator/model-name id. Catalog grew from 3 to 5 models between 2026-08-11 and 2026-08-14; by 2026-08-25 the docs renamed the product “Router API” and added a “private preview” banner — “Email [email protected] to request access” — replacing the earlier no-signup self-serve access.
• Sonar API sunset banner — As of 2026-08-14, the Sonar API pricing docs display: “Sonar Chat Completions is now Agent API. Sonar will be supported until September 27, 2026. Migration guide here.” The migration banner itself predates this check; the explicit end-of-support date is new as of 2026-08-14.
• Consumer and Enterprise pricing pages both currently blocked — perplexity.ai/pricing and perplexity.ai/pro remain behind a Cloudflare bot-verification challenge as of 2026-08-25 (checked with both default and stealth browser identities), so no first-party marketing surface currently shows consumer Pro/Max prices. perplexity.ai/enterprise/pricing briefly came back online on 2026-08-14 with a redesigned page (Individual/Enterprise toggle, three pricing cards, annual-billed per-seat-per-month pricing) after a month blocked, but reverted to a Cloudflare challenge again by 2026-08-25 — the Enterprise seat rates in this article are carried forward from the 2026-08-14 check.

Strategic wins : Why Perplexity AI’s pricing decisions worked

1. The $20 multi-model bundle broke the per-vendor subscription model

By pricing Pro at $20/month with access to every major frontier model, Perplexity made its subscription more cost-efficient than OpenAI, Anthropic, or Google individually. Researchers and power users who previously paid $20 to ChatGPT Plus and $20 to Claude Pro found a single subscription that covered both use cases. This multi-vendor access as the value metric is unique in the market and created a defensible moat that pure-play LLM providers cannot easily replicate. Perplexity’s $34M 2024 revenue and $200M ARR by late 2025 validated the strategy.

2. Generous free tier as the primary acquisition channel

Unlike most AI tools that restrict free tiers to unusable demonstration experiences, Perplexity’s Free plan delivers genuine utility — unlimited basic searches with source citations. This PLG model allowed Perplexity to grow to hundreds of millions of queries per month before investing in paid marketing. The free cap — now 3 Pro Searches per day and a single Research query per month — creates natural conversion moments when users hit limits during high-stakes research sessions, the optimal upgrade trigger.

3. Matching ChatGPT Pro at $200/month signals category intent

The Max plan’s $200/month price point is not arbitrary — it is identical to ChatGPT Pro. This deliberate anchoring communicates that Perplexity is not positioning Max as a niche premium add-on but as a full replacement for the market leader’s highest tier. For enterprise buyers evaluating AI assistant consolidation, the price parity removes the “but it’s cheaper for a reason” objection. See AI pricing strategy for why anchor parity has become a common enterprise positioning tactic.

4. Publisher revenue-share converted adversaries into distribution partners

The July 2024 revenue-share program was a direct response to legal and reputational pressure, but it also functioned as a distribution play: publishers who participate in the program have an incentive to promote Perplexity to their readers as a source of cited AI answers. The $42.5M payout pool announced in January 2026 gives the program credibility at scale. This is a rare example of monetization addressing copyright risk while creating growth levers simultaneously.

5. Zero-markup model resale is the fixed point; tool-wrapper fees are still being tuned

The Agent API passes third-party model tokens through “at direct provider rates with no markup” — a roster that widened on 2026-07-21 from OpenAI, Anthropic, Google, and xAI to also include Z.AI, Moonshot AI, and NVIDIA — and that zero-markup commitment has held through every subsequent change, including the 2026-08-14 update. The tool-wrapper fees layered on top of that resale, by contrast, are visibly still being calibrated: web_search and fetch_url were both cut in half on 2026-07-29 (to $0.0025 and $0.00025), but by 2026-08-14 fetch_url had reverted to its original $0.0005 while web_search held at the cut rate. Read together, the pattern is not a one-way ratchet toward cheaper tools — it’s Perplexity treating the wrapper-fee layer as the place it actively price-tests, while ring-fencing the model-resale margin (zero, permanently) as the part of the offer that stays fixed. That’s a more disciplined value-metric split than it first appears: never compete on model markup, but keep adjusting the price of the retrieval and execution layer Perplexity actually owns until it finds the right number.

Areas to improve : Gaps in Perplexity AI’s pricing approach

1. The 50-seat SCIM cliff is gone, but so is self-serve governance entirely

The 2026-08-14 enterprise page redesign closed the old seat-based compliance gap: SCIM, audit logs, RBAC, and data retention are now marked “Contract required” on both Enterprise Pro ($34/seat/mo) and Enterprise Max ($271/seat/mo) rather than unlocking automatically at 50 seats (or with one Enterprise Max user). That fixes the specific unfairness of a volume threshold, but it trades it for a blanket regression: no team, of any size, can self-serve its way to governance features anymore — a 5-person compliance-sensitive team and a 500-person one both now need a signed contract before they can turn on SCIM. The fix is to publish what “Contract required” actually means in practice — an expected turnaround time, a minimum deal size, or a self-serve path for at least the lighter governance features (audit logs, RBAC) while reserving contracts for deeper integrations (SCIM, custom retention). Right now the page states the rule without stating the cost of complying with it. This is a meaningful pricing architecture problem that likely still costs Perplexity SMB enterprise conversions, just for a different reason than before. See how Intercom handles tiered governance features for a more graduated approach.

2. Each new capability arrives as a new billing unit, and nobody consolidates them

Developers building on Sonar already face two independent cost dimensions — token consumption and per-request search fees — that do not scale proportionally: a query returning a long, heavily-cited answer triggers both large output token costs ($15/1M on Sonar Pro) and a high request fee (up to $14/1K at high search-context depth). The 2026-07-21 sandbox tool adds a third unit, billed per 20-minute container session, so an agent that searches, fetches, and runs code now spends across tokens, invocations, and elapsed time simultaneously. This cost unpredictability compounds with every capability shipped. The fix is a bundled per-outcome rate — a single “Sonar Pro query” or “agent run” price folding tokens, tool calls, and session time into one number, with the itemised meters kept as a breakdown rather than the headline.

3. No predictive spend alerts or budget caps on the API

The developer console shows historical usage but does not offer spend-cap configuration or threshold alerts. A developer who accidentally triggers a deep research loop on Sonar Deep Research could incur $100+ in unexpected charges before noticing, and the new per-session sandbox meter makes that worse: an agent loop that opens a fresh container per iteration bills $0.03 each time regardless of how little of the 20-minute window it uses. Adding configurable spending limits and email/webhook alerts — standard in cloud infrastructure billing (AWS, GCP) — plus a session-reuse guarantee, would reduce developer anxiety and likely increase API adoption from cost-sensitive teams.

4. Every first-party pricing page is dark again — the enterprise recovery didn’t last two weeks

The pattern that looked like a partial fix has fully reverted. Since 2026-07-10, perplexity.ai/pricing has returned a hard 404, and perplexity.ai/pro — briefly reachable on 2026-07-29 — went back behind a Cloudflare challenge by 2026-08-11 and has stayed blocked since. perplexity.ai/enterprise/pricing came back online on 2026-08-14 after sitting blocked since at least 2026-07-21, restoring self-serve verification for exactly the plans procurement teams need to check — but by 2026-08-25 it too reverted to a Cloudflare challenge, undoing that fix inside of 11 days. The net effect: as of 2026-08-25, there is no first-party marketing URL anywhere on perplexity.ai where a prospect — consumer or enterprise — can independently verify a price without already being a customer. Four URL-level incidents in under two months, split evenly between outages and unexplained recoveries, none announced, and now none of the three currently render. The fix is the same one this section has flagged before, made more urgent by the regression: restore a single canonical, crawlable pricing URL and treat page uptime as part of the pricing product rather than an afterthought, since public price verifiability is a precondition for the self-serve motion the rest of Perplexity’s pricing is built around — and right now that precondition holds for neither the enterprise buyer nor the individual one.

5. A GA API pulled back to private preview within two weeks, with no migration notice

Perplexity’s Router API (launched self-serve as the “Gateway API” on 2026-08-11, catalog expanded on 2026-08-14) required an email request to access by 2026-08-25 — the first Perplexity developer product to move from general availability to a gated preview rather than the reverse. Nothing in the docs explains why, and there’s no visible grace period for developers who wired up API keys during the two-week self-serve window: unlike the explicit Sonar sunset notice added the same week (“supported until September 27, 2026”), the Router API’s private-preview banner gives no reason and no reopening timeline. For a company whose docs are otherwise unusually thorough — down to worked cost examples that let this article corroborate rate changes independently — the silence around this specific reversal stands out, and it creates a predictability problem of its own: a developer evaluating any Perplexity API today has to price in the risk that self-serve access itself, not just the rate, can disappear without notice. The fix mirrors the Sonar precedent: publish the reason (capacity, abuse, a pricing-model rework) and a timeline for reopening self-serve, rather than leaving the banner as the only explanation.

Key takeaways

1. Multi-vendor access as the value metric beats feature-per-tier pricing. Perplexity’s decision to bundle GPT, Claude, and Gemini access into one $20 plan created a pricing structure that no single-vendor competitor can replicate. The Agent API rate card shows the corollary: Perplexity resells seven vendors’ models at cost and meters the retrieval and sandbox layer instead, so aggregation only works as a business if you own a meter the vendors you aggregate do not — the Gateway API, launched 2026-08-11 and expanded to 5 models by 2026-08-14, is the sharpest version of that logic yet, selling access to models Perplexity hosts itself at rates it sets, rather than aggregating at cost alone. That mechanic held for barely two weeks: by 2026-08-25 the product — renamed Router API — was pulled behind a private-preview email gate, so today it’s a monetization idea validated at the price-card level only, not yet a live revenue line self-serve developers can act on.

2. Generous free tiers build compounding acquisition advantages. Perplexity’s functional free plan drove hundreds of millions of monthly queries before significant paid marketing. For B2C and prosumer AI products, a free tier that delivers real utility is more efficient than performance marketing at scale. PLG strategy is not just about conversion — it is about organic distribution.

3. Price parity with the market leader signals category ambition, not positioning timidity. Setting Max at exactly $200/month — the same as ChatGPT Pro — is a deliberate market signal. Pricing teams launching premium tiers should consider whether matching rather than undercutting the category leader better serves their positioning goals.

4. Copyright and content provenance are hidden costs in AI search products. Perplexity’s $42.5M publisher payout pool and ongoing litigation represent a structural cost that has no equivalent in pure AI chatbot products. For any product that surfaces web content, the revenue-share or licensing cost of that content must be modeled into unit economics from day one.

5. Removing a seat-threshold gate doesn’t automatically make compliance features more accessible. Perplexity’s 2026-08-14 enterprise redesign replaced its 50-seat SCIM cliff with a blanket “Contract required” label on both Enterprise tiers — a fairer rule, since it no longer punishes teams for being small, but a stricter one, since it removes the volume-based self-serve path entirely. Pricing teams retiring a threshold gate should replace it with a published, priced add-on rather than a sales-contract requirement, or they trade one friction point for a bigger one. See designing enterprise pricing tiers for frameworks.

UBP implications

1. Multi-dimension API billing (tokens + requests + session time) represents the next evolution of AI usage pricing. As AI products incorporate real-time web retrieval, vector search, and external tool calls, pure token-based billing undercounts the true resource cost. The 2026-07-21 sandbox meter pushes this a step further — once an API executes code on the customer’s behalf, the resource being sold is time held in a container ($0.03 per 20-minute session), which neither a token nor a request counter can express. Aggregating usage across dimensions is the core challenge, and agentic APIs will need at least one time-based meter alongside their volume-based ones.

2. The subscription + API hybrid covers more of the customer journey than pure-play models. Perplexity monetizes both the end-user (subscription) and the developer (Sonar API), creating two separate revenue streams from the same underlying model infrastructure. For AI platform companies, this dual-surface pricing strategy has strong revenue forecasting advantages — subscription provides predictable base revenue; API usage provides expansion upside.

3. Publisher revenue-share creates a novel UBP vector: paying for content based on citation volume. Comet Plus ($5/mo, 80% passed to publishers based on citation attribution) is a usage-based payment model applied to content licensing rather than infrastructure. As AI products increasingly depend on third-party content, this citation-based micro-licensing model may become a template for the next generation of content-creator monetization in AI-powered products.

Sources

• Perplexity AI pricing page (accessed 2026-05-29)
• Perplexity Enterprise pricing page (accessed 2026-05-29)
• Perplexity API (Sonar) pricing docs (accessed 2026-05-29)
• Perplexity API (Sonar) pricing docs — current — Agent API web_search/fetch_url tool rates halved on 2026-07-29 (accessed 2026-07-29)
• Perplexity Pro landing page — Pro/Max annual-equivalent monthly pricing and signup credit bonuses (accessed 2026-07-29)
• Perplexity Help Center — What is Perplexity Pro? — current model roster (GPT-5.2, Claude Sonnet 4.6, Gemini 3.1 Pro) (accessed 2026-07-29)
• Perplexity API platform resources — Meet New Sonar (accessed 2026-05-29)
• Perplexity blog — Introducing Perplexity Max (accessed 2026-05-29)
• Perplexity blog — Introducing pplx-api (accessed 2026-05-29)
• Perplexity blog — Introducing the Sonar Pro API (accessed 2026-05-29)
• Perplexity blog — Introducing Comet Plus (accessed 2026-05-29)
• Perplexity Help Center — Subscription and Billing (accessed 2026-05-29)
• Perplexity Help Center — SCIM-based user account management (accessed 2026-07-21)
• Perplexity Help Center — Which Perplexity Subscription Plan is right for you? — plan comparison; free-tier Pro Search and Research limits, Enterprise Pro/Max weekly and monthly quotas, now split into distinct Research Query and Browser Agent (Comet Assistant) Query meters (accessed 2026-07-29)
• Perplexity Help Center — Perplexity Max — “$200/monthly”, “$2000/annually” (accessed 2026-07-29)
• Perplexity Help Center — Enterprise Pricing and Billing FAQ — Enterprise Pro “$40 per month or $400 per year” (now quoted as “a 20% discount versus monthly”), Enterprise Max “$325 per month or $3,250 per year”, new published Education/Nonprofit Enterprise Pro rate (“$30/seat per month or $300/seat annually”) (accessed 2026-07-29)
• Perplexity Help Center — What is Education Pro? — “$10/month (50% off)” (accessed 2026-07-21)
• Perplexity Help Center — How Credits Work on Perplexity — “100 credits equals $1” and per-plan monthly credit allocations (accessed 2026-07-21)
• Perplexity Help Center — API Payment and billing — API credits are purchase-only and separate from Computer credits (accessed 2026-07-21)
• Perplexity API changelog (accessed 2026-07-21)
• Perplexity Enterprise pricing page — redesigned — back online 2026-08-14 after a Cloudflare challenge since at least 2026-07-21; Enterprise Pro “$34/month per seat when billed annually,” Enterprise Max “$271/month per seat when billed annually,” new “Additional security and governance” custom-priced tier, SCIM/audit logs/RBAC/data retention marked “Contract required” for both tiers (accessed 2026-08-14)
• Perplexity API docs — Pricing (current) — Sonar sunset notice (“supported until September 27, 2026”) and Agent API fetch_url price reverted to $0.0005/invocation (accessed 2026-08-14)
• Perplexity API docs — Gateway Models & Pricing — catalog expanded from 3 to 5 models, adding perplexity/nemotron-3.5-lightning-30b-a3b and perplexity/nemotron-3-ultra-550b-a55b (accessed 2026-08-14)
• Perplexity API docs — Router Models & Pricing — same URL, renamed from “Gateway API” to “Router API” throughout the sidebar, page title, and code samples, now behind a “private preview” banner requiring an email request; model catalog and all 5 prices unchanged from the 2026-08-14 check (accessed 2026-08-25)
• Perplexity Enterprise pricing page — reverted to a Cloudflare bot-verification challenge again, 11 days after the 2026-08-14 redesign went live; $34/$271 seat rates carried forward, unverified since (accessed 2026-08-25)

Bottom line

Perplexity AI has built the most efficient consumer AI subscription in the market: $20/month for access to every major frontier model, real-time web search, and cited answers. Its developer APIs extend that infrastructure on a pure-usage model: the Agent API resells seven vendors’ models at zero margin, permanently, while actively tuning the retrieval and code-execution fees around them — fetch_url was halved on 2026-07-29 and reverted on 2026-08-14. The pricing architecture is clever but not flawless: governance features on both Enterprise tiers now require a signed contract rather than unlocking by seat count, the API is metered across three unrelated units (tokens, requests, and per-session container time) plus a fourth — the Router API’s own token rates — that most developers can no longer reach after its 2026-08-25 pull-back to private preview, the Sonar API has a fixed sunset date of September 27, 2026, the search-citation model carries a structural copyright cost, and the consumer pricing page has simply 404’d since 2026-07-10 — the enterprise page briefly came back online the day the governance rules changed, then went dark again itself by 2026-08-25, leaving no first-party pricing URL on the domain reachable at all. But with $200M ARR, a $21B valuation, and the fastest subscription growth in the AI tools category, Perplexity’s pricing strategy has so far out-executed its gaps.

Browse the full pricing blueprint to compare Perplexity AI against other AI platforms.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Gateway API Renamed "Router API" and Gated to Private Preview

Aug 2026

Perplexity's developer docs renamed the Gateway API to "Router API" throughout the sidebar, page titles, and code samples, and added a banner reading "Router API is in private preview. Email [email protected] to request access." The URL (docs.perplexity.ai/docs/gateway/models) and all 5 model prices (deepseek-v4-flash-0731, kimi-k3, glm-5.2, nemotron-3.5-lightning-30b-a3b, nemotron-3-ultra-550b-a55b) are unchanged from the 2026-08-14 catalog, but the product is no longer self-serve — access now requires an email request rather than an API key alone. Confirmed both via Playwright capture and a direct curl of the page's server-rendered source. Separately, perplexity.ai/enterprise/pricing, reachable on 2026-08-14, reverted to a Cloudflare challenge by 2026-08-25 — a capture-access regression on the Enterprise price page, not a confirmed price change.

Agent API fetch_url Price Reverts to $0.0005 — Reverses the 2026-07-29 Cut

Aug 2026

Perplexity's Agent API `fetch_url` tool price doubled back to $0.0005 per invocation, undoing the halving to $0.00025 that took effect on 2026-07-29. Confirmed by two independent worked cost examples on the docs page: the "Agent API Research Preset" total is $0.007 on 2026-08-14 versus $0.00675 on 2026-08-11 — a difference that only reconciles if fetch_url returned to $0.0005. `web_search` ($0.0025), `people_search`/`finance_search` ($0.005), and the $0.03 `sandbox` per-session fee are all unchanged, as are Sonar, Search API, Gateway, and Embeddings rates.

Sonar API Sunset Date Announced — September 27, 2026

Aug 2026

The developer docs' Sonar API pricing page added an explicit sunset date to its existing migration banner: "Sonar Chat Completions is now Agent API. Sonar will be supported until September 27, 2026. Migration guide here." The prior banner (seen through 2026-08-11) named the successor product but carried no end-of-support date. No Sonar token, request-fee, Search API, or Embeddings rate changed alongside this notice — this is a deprecation timeline, not a repricing.

Enterprise Pricing Page Relaunched — Governance Features Restructured

Aug 2026

`perplexity.ai/enterprise/pricing` came back online (200, both default and stealth browser identity) after sitting behind a Cloudflare challenge since at least 2026-07-21. The redesigned page shows Enterprise Pro at $34/seat/month and Enterprise Max at $271/seat/month, both billed annually with no monthly-only price displayed — consistent with the prior $400/seat/yr and $3,250/seat/yr annual totals within rounding. The bigger change is packaging: SCIM, audit logs, role-based access control, custom data retention, and a separate enterprise domain are now marked "Contract required" for BOTH Enterprise Pro and Enterprise Max, replacing the previous 50-seat (or 1 Enterprise Max user) threshold gate that applied only to Enterprise Pro. A new third card, "Additional security and governance" (custom pricing), bundles these contract-gated features plus volume discounts and forward-deployed support. The detailed comparison table also dropped the old absolute Pro-Search/Research/Browser-Agent weekly and monthly quotas in favor of multipliers over the base Pro plan (e.g., file uploads 30x Enterprise Pro / 100x Enterprise Max; asset generation and Comet Agent 2x / 20x).

Gateway API Catalog Expands to 5 Models — 2 NVIDIA Nemotron Additions

Aug 2026

The Gateway API model catalog, launched 2026-08-11 with 3 Perplexity-hosted models (DeepSeek, Kimi, GLM), grew to 5 by 2026-08-14 with two NVIDIA-origin additions: perplexity/nemotron-3.5-lightning-30b-a3b ($0.0115 input / $0.17 output / $0.00115 cache-read per 1M tokens — now the cheapest model in the catalog) and perplexity/nemotron-3-ultra-550b-a55b ($0.25 / $2.50 / $0.25). The 3 original models' prices are unchanged.

Gateway API Launched — Perplexity-Hosted Open-Weight Models

Aug 2026

Perplexity's developer docs added a fifth API surface, the Gateway API, offering "unified access to open-weight models hosted by Perplexity through a single endpoint and a single API key" via OpenAI-compatible Chat Completions and Anthropic-compatible Messages formats. Unlike the Agent API's zero-markup third-party resale, Gateway models are hosted and priced by Perplexity itself: perplexity/deepseek-v4-flash-0731 ($0.13 input / $0.26 output / $0.028 cache-read per 1M tokens), perplexity/kimi-k3 ($3.00 / $15.00 / $0.30), and perplexity/glm-5.2 ($1.40 / $4.40 / $0.14), billed per token with no per-request fees. No existing Sonar, Search API, Agent API tool, or Embeddings rate moved.

Agent API web_search and fetch_url Tool Prices Halved

Jul 2026

Perplexity cut its Agent API web_search tool price in half, to $0.0025 per invocation from $0.005, and fetch_url to $0.00025 from $0.0005. The same cut flowed through to SDK search queries issued inside the sandbox tool, now $0.0025 per request and documented as "same as web_search." people_search, finance_search, and the $0.03 sandbox per-session fee were unchanged, as were Sonar, Search API, and Embeddings rates — corroborated by the docs' own worked cost examples, which total to the new lower figures.

Agent API Adds a Per-Session Sandbox Meter

Jul 2026

Perplexity added a fifth Agent API tool — sandbox, "an isolated container for executing code during an Agent API request" — at $0.03 per container session. It is the first Perplexity meter that counts neither tokens nor requests: a session is a 20-minute billing window rather than a runtime cap, and SDK search queries issued inside it bill separately at $0.005 per request. Zero-markup third-party model resale widened from OpenAI, Anthropic, Google, and xAI to also include Z.AI, Moonshot AI, and NVIDIA. No existing rate moved: Sonar token pricing, request fees by search-context size, the $5.00-per-1,000 Search API, and Embeddings all held.

Comet Browser Goes Free; Comet Plus Add-On at $5/mo

Mar 2026

Perplexity dropped the paywall on its Comet AI browser (iOS, Android, Windows, macOS). A $5/month Comet Plus add-on unlocks premium publisher content inside Comet's AI answers, creating a new micro-subscription layer atop existing plans.

Two-Tier Enterprise — Enterprise Pro + Enterprise Max

Jan 2026

By January 2026 the enterprise page presented a Free / Enterprise Pro / Enterprise Max three-column matrix, confirming a two-tier B2B structure (Enterprise Pro plus a higher Enterprise Max compliance tier) above the consumer plans.

API Restructured — Search API + Agentic Research

Jan 2026

By January 2026 the developer docs split pricing into a standalone Search API ($5.00 per 1,000 requests, "raw web search results with advanced filtering" and no token costs) and an Agentic Research API offering third-party models from OpenAI, Anthropic, Google, and xAI at "direct provider rates with no markup." This separated pure web-search billing from grounded-LLM token billing.

Max Plan Launched at $200/month

Jul 2025

Perplexity launched the Max plan at $200/month (or $2,000/year), matching ChatGPT Pro pricing. Max includes unlimited Labs access, Perplexity Computer (19-model orchestration), early access to Comet browser, and priority access to the latest frontier models. Reported by TechCrunch on 2025-07-02.

Dedicated Enterprise Pricing Page

May 2025

By May 2025 Perplexity ran a standalone enterprise pricing page ("Give your team their time back") with a three-column Free / Pro / Enterprise Pro feature matrix spanning Platform, Security, and Pricing rows, plus a "Have a large team?" contact-sales path — formalizing the seat-based B2B motion alongside the consumer plans.

Sonar API Launched — Search-Native API

Jan 2025

Perplexity replaced pplx-api with the Sonar API: search-native, with live web retrieval, citations, and two tiers (Sonar at $1/1M tokens; Sonar Pro at $3 input/$15 output per million tokens). Per-request search fees added on top of token costs.

Publisher Revenue-Share Program Launched

Jul 2024

Following plagiarism accusations from Forbes, Wired, The New York Times, and others, Perplexity launched a revenue-sharing program for publishers. Partners receive a percentage of ad revenue when their articles are cited in answers. Fortune, Time, Entrepreneur, and 2,400+ others joined by Q1 2026.

Free vs Pro Tiering Formalized

May 2024

By May 2024 the Pro page presented an explicit two-column Free vs Professional split. Free ("Standard") offered unlimited Quick searches but capped Pro Searches at 5 per day with the standard model only. Professional ($20/mo) raised Pro Searches to 600/day, added model selection (GPT-4, Claude 3, Sonar Large/LLaMA 3), up to 50 image answers per day (DALL-E, SDXL), unlimited file uploads, and the $5/month API credit.

Pro Adds $5/mo pplx-api Credits

Jan 2024

By January 2024 the Pro page (still $20/mo, $200/yr) advertised "Over 300 daily" Copilot queries, multimodal file uploads, AI model selection across GPT-4, Claude 2.1, Gemini, and a $5 monthly pplx-api credit bundled into the subscription — tying the consumer plan to the developer API.

pplx-api Launched (Developer API)

Oct 2023

Perplexity introduced pplx-api, providing access to open-source models (Mistral, Llama-based) via REST API. Pricing was usage-based per token. This was a model-inference API, not yet a search-native API.

Pro Plan Live at $20/month

Jun 2023

By June 2023, Perplexity's first paid subscription was live: Pro at $20/month or $200/year (Save $40 a year). The Wayback snapshot shows Pro unlocking GPT-4 access ("Choose GPT-4 for all questions"), 300+ Copilot uses per day, future Pro features at no extra cost, and a private Discord community. A 7-day free trial was offered on the annual plan.

Series A ($26M) — iOS App Launch

Apr 2023

Perplexity raised a $26M Series A from NEA, Databricks Ventures, and others, and launched its iOS app. Company valued at $121M. Product remained fully free.

Perplexity AI Launched

Aug 2022

Perplexity AI launched publicly as a free AI-native answer engine. No paid plans at launch; all search and answer features available free.