specificationVersion: "0.1"
id: serper-rate-limits
name: Serper Rate Limits
generated: '2026-08-13'
method: searched
source: >-
  https://serper.dev (pricing table + FAQ) and Serper's own playground bundles
  https://serper.dev/_next/static/chunks/pages/playground-7ce8960e9fe2fc99.js and
  https://serper.dev/_next/static/chunks/2953-0ed5944c92d83408.js — read 2026-08-13
description: >-
  Serper enforces two independent limits: a queries-per-second concurrency ceiling tied
  to the credit package you bought, and a credit balance that stops accepting queries at
  zero. Credits are deducted only for successful responses. Crucially, a query is not
  always one credit — the cost varies by search type, by result count, and by whether you
  use mini-batch. Serper publishes NO rate-limit response headers, so an agent cannot
  read its remaining budget or a retry delay from the wire.
url: https://serper.dev
limit_count: 4

correction:
  date: '2026-08-13'
  detail: >-
    The previous version of this file listed 50/100/150/200/300 QPS for
    Free/Starter/Standard/Scale/Ultimate. Serper's published pricing table states
    50 (Starter), 100 (Standard), 200 (Scale), 300 (Ultimate) and publishes no QPS figure
    for the free trial. Corrected below.

headers:
  published: false
  request_headers: []
  response_headers: []
  retry_after: false
  status_on_exhaustion: 429
  detail: >-
    No RateLimit-*, X-RateLimit-* or Retry-After header is documented, and none was
    observed. The only runtime signal is the HTTP status and the JSON body
    {"message": ..., "statusCode": ...}. Remaining credit balance is visible only in the
    Serper dashboard, not in an API response — except on the webpage scrape surface,
    where Serper states the credits consumed by that call are returned in the response
    body.

rateLimits:
  - name: Starter QPS Limit
    description: Concurrency ceiling for the Starter credit package.
    scope: per-account
    plan: Starter
    limit: 50
    period: second
    unit: requests
    notes: 50,000 credits at $1.00/1k. Credits valid 6 months.

  - name: Standard QPS Limit
    description: Concurrency ceiling for the Standard credit package.
    scope: per-account
    plan: Standard
    limit: 100
    period: second
    unit: requests
    notes: 500,000 credits at $0.75/1k.

  - name: Scale QPS Limit
    description: Concurrency ceiling for the Scale credit package.
    scope: per-account
    plan: Scale
    limit: 200
    period: second
    unit: requests
    notes: 2,500,000 credits at $0.50/1k.

  - name: Ultimate QPS Limit
    description: >-
      Default concurrency ceiling for the Ultimate credit package. Serper's FAQ states
      this allows roughly 15,000-18,000 searches per minute and that the limit can be
      raised for a specific use case on request.
    scope: per-account
    plan: Ultimate
    limit: 300
    period: second
    unit: requests
    raisableOnRequest: true
    notes: 12,500,000 credits at $0.30/1k.

freeTrial:
  credits: 2500
  requestsPerSecond: null
  note: >-
    Serper publishes 2,500 free credits with no credit card, but states no QPS ceiling
    for the trial. Recorded as unknown rather than guessed.

creditCosts:
  description: >-
    The credit cost of a query, read from the calculator in Serper's own playground. This
    is the real unit of consumption behind the QPS limits and the balance.
  source: >-
    https://serper.dev/_next/static/chunks/pages/playground-7ce8960e9fe2fc99.js
  default: 1
  byType:
    - type: search
      credits: 1
    - type: images
      credits: 1
      multiplier_note: Doubles to 2 when num > 10.
    - type: videos
      credits: 1
    - type: news
      credits: 1
    - type: places
      credits: 1
    - type: maps
      credits: 3
    - type: reviews
      credits: 1
    - type: shopping
      credits: 2
    - type: scholar
      credits: 1
    - type: patents
      credits: 1
    - type: autocomplete
      credits: 1
    - type: lens
      credits: 3
    - type: webpage (scrape.serper.dev)
      credits: "2, 6 or 10"
      note: >-
        Serper's playground states verbatim: "Most webpages typically incur 2 credits,
        but for harder-to-scrape pages that require more advanced scraping techniques,
        the cost can be 6 or 10 credits depending on the difficulty. The number of
        credits used is returned in the response."
  modifiers:
    - name: Mini-batch
      effect: multiply by 3
      detail: >-
        Submitting an array of up to 100 queries to one endpoint costs 3x the single-query
        cost. Not available on the reviews or product-reviews types.
    - name: Large result set on Images
      effect: multiply by 2
      detail: num > 10 on Google Images doubles the cost of the query.
  gated_types_note: >-
    The playground also prices types that are gated behind an @serper.dev account
    (search-full at 3 or 6 credits, bing-*, product-reviews at 2, scholar-case at 2,
    duckduckgo search at 2). Those are not part of the public surface and are recorded
    here only so the pricing table is not misread as incomplete.

globalNotes: >-
  Credits are deducted per successful response; Serper's FAQ states failed queries are
  not charged. When the credit balance reaches zero, Serper stops accepting new queries.
  Typical response time is 1-2 seconds, rising to 2-4 seconds when Serper has to retry
  against Google. Serper does not cache, so every call is a live Google query.