# Provenance stamped 2026-10-07 (roadmap#943): written by an API Evangelist sweep (dated 2026-06-14 by the file itself), not
# harvested from the provider. `method: generated` so the economics campaign may replace it with a
# dated, verbatim transcription; every figure below is unverified until it does.
method: generated
generated: '2026-06-14'
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: GMX
providerId: gmx
created: '2026-06-14'
modified: '2026-06-14'
tags:
  - DeFi
  - Perpetual Exchange
  - DEX
  - Trading
  - Liquidity Pools
  - GM Tokens
  - Arbitrum
  - Avalanche
  - Web3
  - Rate Limiting
  - Quotas
  - Throttling
description: >-
  Scaffolded rate limit definitions for the GMX API surface. GMX does not
  publish explicit rate limit headers or throttling policies in its public
  documentation as of June 2026. The Oracle API (gmxinfra.io) provides
  fallback hosts for high availability. Scaffold values below are conservative
  defaults; replace with provider-published values if and when available.
headers:
  limit: X-RateLimit-Limit
  remaining: X-RateLimit-Remaining
  reset: X-RateLimit-Reset
  retryAfter: Retry-After
  policy: RateLimit-Policy
responseCodes:
  throttled: 429
  quotaExceeded: 429
  serviceUnavailable: 503
limits:
  - tier: public
    name: Oracle API Default
    scope: ip
    metric: requests_per_minute
    limit: 60
    burst: 100
    timeFrame: minute
    applies:
      - https://arbitrum-api.gmxinfra.io
      - https://avalanche-api.gmxinfra.io
    description: >-
      Scaffold default for the public Oracle API. No key required. Use
      documented fallback hosts (arbitrum-api-fallback.gmxinfra.io,
      arbitrum-api-fallback.gmxinfra2.io) if the primary host is unavailable.
  - tier: public
    name: GMX API Default
    scope: ip
    metric: requests_per_minute
    limit: 60
    burst: 100
    timeFrame: minute
    applies:
      - https://arbitrum.gmxapi.io/v1
      - https://avalanche.gmxapi.io/v1
    description: >-
      Scaffold default for the primary GMX API. Two independent peer regions
      per chain (gmxapi.io and gmxapi.ai) provide redundancy.
  - tier: public
    name: GraphQL / Subsquid Default
    scope: ip
    metric: requests_per_minute
    limit: 30
    burst: 60
    timeFrame: minute
    applies:
      - https://gmx.squids.live/gmx-synthetics-arbitrum:prod/api/graphql
    description: >-
      Scaffold default for Subsquid-powered GraphQL queries. Subsquid imposes
      its own fair-use limits; check Subsquid documentation for current quotas.
recoveryStrategies:
  - name: Exponential Backoff
    description: >-
      Wait for the Retry-After header value (or back off exponentially up to
      60 s) before retrying throttled requests.
  - name: Fallback Hosts
    description: >-
      For Oracle API, switch to the documented fallback hosts
      (arbitrum-api-fallback.gmxinfra.io / gmxinfra2.io) on 429 or 503
      before applying backoff.
  - name: Dual Region
    description: >-
      For the GMX API, alternate between the .io and .ai peer-region base
      URLs to distribute load and improve resilience.
notes:
  - GMX has not published explicit rate limit headers or throttling policies
    as of June 2026; scaffold values are conservative defaults.
  - Replace scaffold values with provider-published limits when available.