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: Bifrost
providerId: bifrost
created: '2026-05-04'
# Provenance stamped 2026-08-11: this artifact was written by the API Evangelist
# bulk sweep dated 2026-05-04, not harvested from the provider. See roadmap#35.
method: generated
modified: '2026-05-05'
reconciled: false
tags:
  - AI Gateway
  - LLM
  - Open Source
  - Rate Limiting
description: Bifrost is a self-hosted gateway, so there are no provider-imposed rate limits — limits are
  whatever the operator configures via virtual keys, budgets, and per-key rate-limit policies. Effective
  ceilings are governed by (1) the host's compute/network capacity and (2) the upstream LLM provider's
  own rate limits, which Bifrost surfaces via fallback and load-balancing logic.
sources:
  - https://docs.getbifrost.ai/
  - https://getbifrost.ai/pricing
notes: No published per-second/per-minute rate-limit numbers exist for Bifrost itself. The OSS distribution
  ships virtual-key-level rate limits and budgets that operators configure. Numeric values below are placeholders
  intended only to illustrate the dimensions Bifrost can enforce.
responseCodes:
  throttled: 429
limits:
  - name: Virtual Key Request Rate
    scope: virtual-key
    metric: requests_per_minute
    limit: 'operator-configured'
    notes: Bifrost virtual keys carry rate-limit policy set by the deploying operator.
  - name: Virtual Key Budget
    scope: virtual-key
    metric: spend_per_period
    limit: 'operator-configured'
    notes: Per-key budget policy enforces upstream-cost spend caps.
  - name: Upstream Provider Limits
    scope: provider/model
    metric: varies
    limit: 'inherited from upstream provider (OpenAI, Anthropic, etc.)'
    notes: Bifrost transparently respects and falls back across upstream provider rate limits.
policies:
  - name: Adaptive Load Balancing (Enterprise)
    description: The Enterprise edition redistributes load across upstream providers based on observed
      latency and error rate.
  - name: Automatic Fallback
    description: On 429 or 5xx from an upstream provider, Bifrost retries against a configured fallback
      provider/model.
  - name: Semantic Caching
    description: Cache hits avoid upstream calls entirely, reducing effective rate-limit pressure.
  - name: Operator-Owned Configuration
    description: Because Bifrost is self-hosted, all rate-limit thresholds are owned by the operator —
      not the project.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com