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: Cleanlab
providerId: cleanlab
created: '2026-06-21'
modified: '2026-06-21'
reconciled: false
tags:
- AI
- LLM
- Trustworthiness
- Data Quality
- Guardrails
- Rate Limiting
- Quotas
- Throttling
description: >-
  Cleanlab applies per-account rate limits to its hosted APIs - the Trustworthy
  Language Model (TLM), Codex validation, and Cleanlab Studio inference. TLM
  enforces concurrency and request-rate limits that vary by tier (free vs.
  paid). The Python clients retry transient errors and respect rate limiting.
  Specific per-tier numeric limits are not consistently published and are not
  reconciled in this artifact.
notes: >-
  Verify concurrency and request-rate limits in the relevant Cleanlab console
  (tlm.cleanlab.ai, app.cleanlab.ai, codex.cleanlab.ai) during reconciliation;
  limits change between free, paid, and enterprise tiers.
sources:
- https://help.cleanlab.ai/tlm/faq/
- https://help.cleanlab.ai/tlm/
- https://help.cleanlab.ai/studio/
- https://help.cleanlab.ai/codex/
responseCodes:
  throttled: 429
limits:
- name: TLM Concurrent Requests
  scope: account
  metric: concurrent_requests
  limit: see provider documentation
  notes: Maximum in-flight TLM requests; varies by free vs. paid tier.
- name: TLM Requests Per Minute
  scope: account
  metric: requests
  limit: see provider documentation
  notes: Per-account TLM request rate; varies by tier.
- name: Codex Validation Requests
  scope: project
  metric: requests
  limit: see provider documentation
  notes: Per-project validate request rate.
- name: Studio Inference Requests
  scope: account
  metric: requests
  limit: see provider documentation
  notes: Per-account deployed-model prediction request rate.
policies:
- name: Tiered Limits
  description: Limits raise as accounts move from free to paid usage and via Enterprise agreements.
- name: Backoff Strategy
  description: Clients should implement exponential backoff with jitter and honor Retry-After; the official Python clients retry transient errors automatically.
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com