# Rate limits transcribed from the provider's own documentation — see provenance. Quoted, not derived.
generated: '2026-09-20'
method: searched
source: https://docs.digitalocean.com/products/inference/details/limits/
sources:
- https://docs.digitalocean.com/products/inference/details/limits/
provenance:
  tool: planning/api-economics/harvest_ratelimits.py
  run: 20260920T133210
  grounding: every limit_text was found verbatim in the fetched page text
published: true
note: Page lists tiered quotas (Tier 1-5) for agents, knowledge bases, RPM (120-4,500) and TPM in a table
  that cannot be quoted verbatim as a single string; also batch limits (50,000 requests per file, 200
  MB file size, 24 hour completion window). No rate-limit headers or status codes documented.
limit_count: 6
limits:
- name: Inference Router requests per minute
  limit_text: Routers support 1,000 requests per minute.
  limit: 1000
  window: minute
  scope: endpoint
  endpoint: Inference Router
- name: Batch enqueue token limit
  limit_text: 10 billion tokens per model per account
  limit: 10000000000
  scope: account
  endpoint: Batch Inference
- name: Knowledge bases per team
  limit_text: Each team can create up to 120 knowledge bases.
  limit: 120
  scope: account
  endpoint: Knowledge Bases
- name: Presets per team
  limit_text: Each team can create up to 100 presets.
  limit: 100
  scope: account
  endpoint: Presets
- name: Custom metrics per team
  limit_text: Each team can create up to 100 custom metrics.
  limit: 100
  scope: account
  endpoint: Custom Metrics
- name: Web crawler pages
  limit_text: the crawler indexes up to 5,500 pages
  limit: 5500
  scope: endpoint
  endpoint: Web crawling data sources