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: Midjourney providerId: midjourney 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: - Rate Limiting - AI - Image Generation description: Midjourney does not currently expose a public, metered HTTP API. Generations are scoped to Discord and the web app and are governed primarily by per-plan Fast GPU minute budgets and concurrent-job caps rather than per-second rate limits. Concurrency limits and Relax-mode queue behavior vary by tier. sources: - https://docs.midjourney.com notes: docs.midjourney.com refused automated fetches (403). Reconciliation deferred — confirm per-tier concurrent job ceilings and any queue / cooldown rules from the live plan page. responseCodes: {} limits: - name: Fast GPU minutes (monthly) scope: account metric: gpu-minute limit: 'Per-plan; see plan page' timeFrame: month - name: Concurrent jobs (Fast) scope: account metric: concurrent_requests limit: 'Per-plan; typically 3 (Basic) up to 12+ (Mega) — verify on live docs' - name: Relax mode queue scope: account metric: varies limit: 'Standard / Pro / Mega only; queued behind Fast jobs, no per-minute SLA' policies: - name: Plan-bound concurrency description: Each tier has a fixed maximum of concurrent Fast jobs; additional generations queue until a slot frees. - name: Relax mode degradation description: Once Fast minutes are exhausted, supported tiers fall back to Relax mode (queued) until the next billing cycle or Fast top-up. - name: Stealth gating description: Stealth Mode is a Pro / Mega entitlement, not a rate limit, but affects which generations are publicly visible. maintainers: []