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: CentML
providerId: centml
created: '2026-06-21'
modified: '2026-06-21'
reconciled: false
tags:
- AI
- LLM
- Inference
- Serverless
- GPU
- Rate Limiting
- Quotas
- Throttling
description: >-
  CentML's serverless inference API enforces per-account rate limits expressed
  as requests per minute and tokens per minute, which vary by model and account
  tier. Dedicated deployments are governed by the capacity of the provisioned
  GPU hardware and configured autoscaling (min/max replicas) rather than shared
  account-level token limits. Specific per-model limit values are not reconciled
  in this artifact.
notes: >-
  Verify per-model and per-account limits in the CentML platform / documentation
  on reconciliation; serverless limits change as accounts move between tiers, and
  dedicated throughput depends on the selected hardware instance and replica count.
sources:
- https://docs.centml.ai/apps/serverless
- https://docs.centml.ai/apps/inference
- https://centml.ai/pricing/
responseCodes:
  throttled: 429
limits:
- name: Requests Per Minute (RPM)
  scope: account
  metric: requests
  limit: see provider documentation
  notes: Per-model RPM on serverless endpoints, varies by tier and model.
- name: Tokens Per Minute (TPM)
  scope: account
  metric: tokens
  limit: see provider documentation
  notes: Per-model TPM on serverless endpoints, varies by tier and model.
- name: Concurrent Requests
  scope: account
  metric: requests
  limit: see provider documentation
  notes: Concurrency permitted against serverless endpoints, varies by tier.
- name: Dedicated Replica Capacity
  scope: deployment
  metric: replicas
  limit: configured via min/max replicas
  notes: Dedicated deployment throughput is bounded by provisioned GPU hardware and autoscaling settings, not shared account token limits.
policies:
- name: Tiered Limits
  description: Limits raise as accounts move from free / trial to paid usage and via Enterprise agreements.
- name: Backoff Strategy
  description: Clients should implement exponential backoff with jitter and honor Retry-After on 429 responses.
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com