Artificial IndexVersion 1.1.0 · Humanity's AI Index
deep-verified profile

Perplexity

AI answer engine focused on retrieval-assisted research.

ProviderPerplexity AI
Catalog profileAvailable
Evidence completeness76%
LifecycleActive
Checked2026-09-07

Core record

Status: Generally available

Access: Web, desktop, mobile, enterprise plans and separately billed public APIs

API status: Public APIs available; Router API is private preview

Openness: Proprietary service; Router also hosts third-party open-source models

Modalities: Text, web search, image/file input, image/video results

Best suited for

Answer-engine and research-oriented discovery workflows.

Limitations

Exact context and output limits depend on the selected model or surface. Regional availability and benchmark evidence remain pending.

API / auth

Interface: Agent, Search, Sonar, Embeddings and private-preview Router APIs

Base URL: https://api.perplexity.ai

Authentication: API key with Bearer authentication

Compatibility: Official Python/TypeScript SDKs and documented OpenAI-compatible endpoints

Rate limits: Published by API and usage tier; rejected 429 requests are not billed

Deployment

Hosted webDesktopMobileHosted APIsRouter: private preview

Models / versions

ModelContextMax outputModalitiesLicenseStatus
SonarPending exact official limitPending exact official limitText + grounded searchProprietary serviceCurrent
Sonar ProPending exact official limitPending exact official limitText and mediaProprietary serviceCurrent
Sonar Reasoning ProPending exact official limitPending exact official limitText + grounded reasoningProprietary serviceCurrent
Sonar Deep ResearchPending exact official limitPending exact official limitResearch output with citationsProprietary serviceCurrent

API pricing snapshot

SKU / modelUnitInputOutputFlatNotes
SonarUSD / 1M tokens$1$1—Plus $5/$8/$12 per 1K requests by search context.
Sonar ProUSD / 1M tokens$3$15—Plus $6/$10/$14 per 1K requests; Pro Search can cost more.
Sonar Reasoning ProUSD / 1M tokens$2$8—Plus $6/$10/$14 per 1K requests.
Sonar Deep ResearchUSD / 1M tokens$2$8—Separate citation, search-query and reasoning charges.
Search APIUSD / 1K requests——$5No additional token charge.

Integrations

Google DriveNotionAsanaJiraConfluenceMicrosoft 365TeamsSlackBoxDropboxGitHubSnowflakeMCP

Regions

Pending authoritative regional availability list. No country coverage is inferred.

Governance / privacy

Official API documentation states zero content retention for the Chat Completions API, excludes prompt/response content from billing metadata, and lists SOC 2 Type II through the Perplexity Trust Center. These claims are scoped to the API documentation and are not generalized to every consumer surface.

Benchmarks

Pending methodology-aware benchmark audit. Vendor and independent benchmark results remain separate.

Field verification matrix

identity: verifiedorganization: verifiedcategories: verifiedstatus: verifiedaccess: verifiedapi: verifiedauthentication: verifiedmodels: verifiedcontext: pendingmax_output: pendingmodalities: verifiedopenness_license: partialpricing: verifiedrate_limits: verifieddeployment: partialregions: pendingintegrations: verifiedgovernance: verifiedbenchmarks: pendinglifecycle: verifiedsources: verified

Evidence

Official API overviewFields: API, models, access · checked 2026-09-07Confirms the current API families and keeps Router explicitly marked private preview.Source ↗Official API pricingFields: pricing, models · checked 2026-09-07Current official rate card; staged third-party pricing was not used.Source ↗Official rate limitsField: rate limits · checked 2026-09-07Endpoint- and tier-specific limits replace the outdated claim that limits were unpublished.Source ↗Official privacy and security documentationField: governance · checked 2026-09-07Claims are kept within the API scope stated by Perplexity.Source ↗Official connectors collectionField: integrations · checked 2026-09-07Only officially listed integrations are recorded.Source ↗
Evidence completeness is 76%. Context, max output, regions and benchmarks remain pending; deployment and openness/license remain partial. No unsupported claim was promoted.