generated: '2026-09-02'
method: searched
source: https://standardcompute.com/smart-pacing
also:
  - https://standardcompute.com/fair-use
  - https://standardcompute.com/llms-full.txt
  - https://standardcompute.com/status
summary: >-
  Standard Compute deliberately does NOT publish a classic request-rate limit.
  There are no per-minute 429 walls, no 5-hour session windows and no weekly caps.
  The binding constraint is a fixed MONTHLY COMPUTE BUDGET per plan; requests run
  at full speed until it is consumed and then return HTTP 402 until the billing
  period renews. Under extreme sustained load ahead of budget the gateway engages
  "pacing" — requests are spaced out but still succeed — instead of rejecting them.
  Anti-abuse limits exist but their thresholds are not published.
limit_count: 3
rate_limits:
  - scope: per-account (monthly)
    window: calendar billing period (1 month)
    limit: plan compute budget in USD of upstream work — $20 Starter / $41 Economy / $95 Standard / $269 Pro; higher Business tiers stated as included API-work value ($1,800 / $3,700 / $9,300)
    burst: unrestricted within the budget — the provider states pace on day one equals pace on day thirty
    exhaustion_status: 402
    exhaustion_behavior: >-
      Requests stop cleanly until the period renews; upgrading a tier takes effect
      immediately and resumes service.
    source: https://standardcompute.com/fair-use
  - scope: per-account (sustained load)
    window: continuous
    limit: not published — engages only under "extreme sustained load" running far ahead of the monthly budget
    behavior: >-
      Smooth pacing: additional latency is inserted between requests. The provider
      explicitly states requests keep succeeding and no 429 is returned, so an agent
      loop does not break.
    exhaustion_status: null
    source: https://standardcompute.com/smart-pacing
  - scope: per-key (authentication)
    window: not published
    limit: not published
    behavior: failed authentication attempts are rate limited to mitigate brute-force key guessing
    exhaustion_status: 401
    source: https://standardcompute.com/security
optional_controls:
  - name: Smart pacing
    default: 'off'
    effect: >-
      Dashboard switch that spreads the remaining monthly budget evenly across the
      month, for always-on agents that should not drain the budget in week one.
    source: https://standardcompute.com/smart-pacing
concurrency:
  starter: one agent at a time
  economy_and_above: parallel agents permitted
  source: https://standardcompute.com/pricing
response_headers:
  published: false
  observed: []
  note: >-
    PROBED 2026-09-02. An unauthenticated POST to /v1/chat/completions returned
    HTTP 401 with no X-RateLimit-*, no RateLimit-* and no Retry-After header. The
    provider documents no runtime rate-limit headers anywhere, and the published
    OpenAPI declares no 402/429 responses. An agent therefore has NO in-band signal
    of remaining budget or of pacing — the only budget view is the human dashboard.
    This is the clearest agent-readiness gap on this API and is captured as a
    recommendation in overlays/standard-compute-openapi-overlay.yaml.
  retry_after: not sent
  status_on_exhaustion: 402