generated: '2026-08-26'
method: searched
source: https://vynaris.com/docs
docs:
- https://vynaris.com/docs#errors
- https://api.vynaris.com/agents.md
- https://vynaris.com/uncensored-models
- https://vynaris.com/terms
limit_count: 0
note: >-
  Vynaris publishes NO rate limits. Every public surface was read on 2026-08-26
  - the docs page, agents.md, llms.txt, the terms, the pricing page and the
  hosted-model pages - and none states a requests-per-second, per-minute,
  per-day, token-per-minute or concurrency ceiling. This is an honest zero, not
  an unchecked field. It is a real gap for an autonomous caller: the provider
  markets programmatic self-onboarding and agent operation, and an agent that
  self-onboards has no published number telling it how hard it may drive the
  API.
rate_limits: []
response_headers:
  published: false
  x_ratelimit: none documented
  ratelimit_rfc: none documented
  retry_after: none documented
  note: >-
    No X-RateLimit-*, RateLimit-* or Retry-After header appears in any
    published Vynaris surface. The documented response headers are all cost
    telemetry (x-vynaris-request-id, x-vynaris-served-model,
    x-vynaris-cost-usd, x-vynaris-baseline-usd, x-vynaris-saved,
    x-vynaris-balance) - the provider instruments spend, not throughput.
status_on_exhaustion:
  http_429: not documented
  note: >-
    No 429 is documented. The documented failure modes are 401, 402 and 502
    only. The terms do refer to "bypass its rate or budget limits" as
    prohibited conduct, which implies limits exist and are enforced - they are
    simply not published.
economic_limits:
  description: >-
    The only backpressure Vynaris documents is financial rather than
    throughput-based, and it is well specified.
  controls:
  - control: prepaid balance floor
    behavior: requests stop when available balance reaches zero
    status: 402
    terminal: true
    source: https://vynaris.com/docs#errors
  - control: per-key spend cap
    behavior: optional cap per API key; hitting it returns the same terminal 402
    source: https://api.vynaris.com/agents.md
  - control: no overages
    behavior: usage is never billed beyond the prepaid balance
    source: https://vynaris.com/pricing
  - control: in-flight float
    behavior: >-
      "balance never goes materially negative - at most one in-flight request
      of float, because streams are never cut mid-completion"
    source: https://api.vynaris.com/agents.md
concurrency_guidance:
  published: true
  quantified: false
  scope: Vynaris-hosted reduced-refusal models only
  text: >-
    "Concurrent requests use shared queued capacity; do not create a new worker
    per end user." The first request after an idle period can take longer while
    private capacity starts and weights load.
  note: >-
    This is directional advice, not a limit. No maximum concurrency, queue
    depth, or cold-start timeout is given.
  source: https://vynaris.com/docs#hosted-models