specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: Amazon Lex
providerId: amazon-lex
generated: '2026-09-17'
method: searched
source: https://docs.aws.amazon.com/lexv2/latest/dg/quotas.html
modified: '2026-09-17'
created: '2026-05-04'
supersedes: >-
  The 2026-05-04 bulk-sweep scaffold, which asserted X-RateLimit-* headers and a generic per-tier quota
  model. Amazon Lex V2 publishes neither. Replaced wholesale with the published quota tables.
tags:
- Quotas
- Throttling
- Service Quotas
description: >-
  Amazon Lex V2 limits are AWS Service Quotas — per AWS account, per Region — not per-plan API rate
  limits. They split into build-time quotas (how large a bot may be) and runtime quotas (how many
  concurrent conversations an alias may carry). Adjustable quotas are raised through the Service Quotas
  console; some require a support case.
console: https://console.aws.amazon.com/servicequotas/home/services/lex/quotas
headers:
  limit: null
  remaining: null
  reset: null
  retryAfter: null
  policy: null
headers_note: >-
  MEASURED ABSENCE. Amazon Lex V2 publishes NO rate-limit response headers — no RateLimit-*, no
  X-RateLimit-*, and no documented Retry-After. An agent gets no advance warning that it is
  approaching a limit; it learns only from the 429. Current consumption is visible out-of-band through
  the Service Quotas API and CloudWatch, never on the response.
responseCodes:
  throttled: 429
  exception: ThrottlingException
  quota_exceeded: 402
  quota_exception: ServiceQuotaExceededException
scope: per AWS account, per AWS Region (some runtime quotas scoped per bot alias)
limit_count: 24
limits:
- name: Bots per AWS account
  type: build-time
  scope: account/region
  limit: 100
  adjustable: true
  self_service: true
- name: Bot channel associations per AWS account
  type: build-time
  scope: account/region
  limit: 5000
  adjustable: false
- name: Parallel locale builds per AWS account
  type: build-time
  scope: account/region
  limit: 5
  adjustable: true
  self_service: false
- name: Versions per bot
  type: build-time
  scope: bot
  limit: 100
  adjustable: false
- name: Intents per locale in each bot
  type: build-time
  scope: bot locale
  limit: 1000
  limit_note: 1,000 in en-AU, en-GB and en-US; 250 in all other locales.
  adjustable: true
  self_service: false
- name: Slots per locale in each bot
  type: build-time
  scope: bot locale
  limit: 4000
  limit_note: 4,000 in en-AU, en-GB and en-US; 2,000 in all other locales.
  adjustable: false
- name: Slots per intent
  type: build-time
  scope: intent
  limit: 100
  adjustable: false
- name: Sample utterances per intent
  type: build-time
  scope: intent
  limit: 1500
  adjustable: true
  self_service: true
- name: Custom slot type values and synonyms per locale
  type: build-time
  scope: bot locale
  limit: 50000
  adjustable: false
- name: Text response length
  type: build-time
  scope: message
  limit: 4000
  unit: characters
  adjustable: false
- name: Size of custom grammar slot type XML file
  type: build-time
  scope: slot type
  limit: 100
  unit: KB
  adjustable: false
- name: Concurrent Automated Chatbot Designer analysis jobs
  type: build-time
  scope: account/region
  limit: 10
  adjustable: false
- name: Input text size for RecognizeText and RecognizeUtterance
  type: runtime
  scope: request
  limit: 1024
  unit: characters
  adjustable: false
- name: Speech input length for RecognizeUtterance
  type: runtime
  scope: request
  limit: 55
  unit: seconds
  adjustable: true
  self_service: false
- name: Size of RecognizeUtterance headers
  type: runtime
  scope: request
  limit: 16
  unit: KB
  adjustable: false
- name: Concurrent text-mode conversations (TestBotAlias)
  type: runtime
  scope: bot alias
  limit: 2
  adjustable: false
  operations: [RecognizeText, RecognizeUtterance, StartConversation]
- name: Concurrent text-mode conversations (other aliases)
  type: runtime
  scope: bot alias
  limit: 50
  adjustable: true
  self_service: false
  operations: [RecognizeText, RecognizeUtterance, StartConversation]
- name: Concurrent voice-mode conversations for RecognizeUtterance (other aliases)
  type: runtime
  scope: bot alias
  limit: 125
  adjustable: true
  self_service: false
- name: Concurrent voice-mode conversations for StartConversation (other aliases)
  type: runtime
  scope: bot alias
  limit: 400
  adjustable: true
  self_service: false
- name: Concurrent session management operations (other aliases)
  type: runtime
  scope: bot alias
  limit: 50
  adjustable: true
  self_service: false
  operations: [PutSession, GetSession, DeleteSession]
- name: Maximum input size to a Lambda function
  type: runtime
  scope: request
  limit: 12
  unit: KB
  adjustable: false
- name: Maximum output size of a Lambda function
  type: runtime
  scope: request
  limit: 50
  unit: KB
  adjustable: false
- name: Maximum timeout of a Lambda function
  type: runtime
  scope: request
  limit: 30
  unit: seconds
  adjustable: true
  self_service: false
- name: Maximum duration for a single conversation
  type: runtime
  scope: session
  limit: 15
  unit: minutes
  adjustable: false
agent_note: >-
  The TestBotAlias caps every concurrency dimension at 2. An agent load-testing against TSTALIASID will
  throttle at two concurrent conversations and conclude the service is far more constrained than it is.