generated: '2026-08-05'
method: searched
source: https://docs.0g.ai/developer-hub/building-on-0g/compute-network/router/rate-limits
api: 0G Compute Router API
summary: >-
  0G publishes the rate-limit MECHANISM but deliberately not the numbers. The docs page states
  outright that "the exact thresholds depend on your account state and may evolve as we tune
  them — this page documents how to observe and react to the limit, not the specific numbers."
  Every inference response therefore carries your current limit in a header, which is the
  intended way to discover it. Recorded here as published: signaling is complete, published
  numeric limits are none.
published_numeric_limits: false
rate_limits:
- name: per-account request limit
  scope: account
  window: minute
  limit_count: null
  limit_note: >-
    Not published. Read X-RateLimit-Limit-Requests off any inference response to learn your
    current per-minute limit.
  discovery: response header
signaling:
  headers:
  - name: X-RateLimit-Limit-Requests
    meaning: your current per-minute request limit
  - name: X-RateLimit-Remaining-Requests
    meaning: how many requests remain in this window
  - name: X-RateLimit-Reset-Requests
    meaning: ISO-8601 timestamp when the window resets
  standard: OpenAI-compatible rate-limit headers
  present_on: every inference response
exceeded:
  status: 429
  header: Retry-After
  header_unit: seconds
  example_value: 15
  body:
    error:
      message: Rate limit exceeded. Please try again later.
      type: rate_limit_error
      code: rate_limit_exceeded
  guidance: >-
    Honor Retry-After. Do not retry in a tight loop — the Router will keep returning 429 and
    real requests will be delayed behind them.
roadmap:
  status: announced, not shipped
  detail: >-
    Per-API-key throughput controls — explicit RPM (requests per minute) and TPM (tokens per
    minute) budgets settable in the dashboard — are documented as "coming soon".
related_limits:
- name: min_cost admission floor
  detail: >-
    An inference request is admitted only if the account can cover a minimum charge. The floor
    rises from 0.00001 0G to 0.01 0G. Below it the request is refused at admission rather than
    rate-limited.
  source: openapi/0g-labs-router-openapi.yml (info.description Changelog)
- name: testnet faucet
  detail: 0.1 0G per day per address on the Galileo testnet faucet.
  source: https://faucet.0g.ai
cross_links:
  errors: errors/0g-labs-problem-types.yml
  conventions: conventions/0g-labs-conventions.yml