Source: OpenAI — Scale Tier and Reserved Tier pages

OpenAI's committed-capacity offerings: Scale Tier token units (30-day minimum, 99.9% uptime SLA) for models before GPT-5.6, and Enterprise Reserved Tier ($/minute reservations) for GPT-5.6 onward.

Contents
  • What: vendor product pages
  • Publisher: OpenAI
  • Date: undated, read 2026-11-14 via search-result extracts (direct fetch returned HTTP 403; verify on the live pages)
  • Original: API Scale Tier and API Reserved Tier

Summary

The two pages describe OpenAI’s direct-API equivalents of provisioned throughput. Scale Tier sells per-minute token units for one model snapshot; Reserved Tier, for GPT-5.6 and later, sells a per-minute dollar budget.

Key claims

  • Scale Tier: buy input/output tokens-per-minute “units” for a specific model snapshot, with a 30-day minimum commitment, prioritized compute and a 99.9% uptime SLA. Older models sold separate input and output units (GPT-4.1: one input unit of 30K input TPM at $110/day); newer ones sell a combined bundle where input, cached input and output draw down at different rates.
  • Scale Tier applies to models released before GPT-5.6; GPT-5.6 onward uses Reserved Tier.
  • Reserved Tier (Enterprise, via sales): reserve a dollar amount per minute for one model. Standard and Fast-mode usage within it is covered; Batch and Flex are excluded; usage above it is billed pay-as-you-go. Purchases raise rate limits to secure capacity at peak.

Feeds

Provisioned throughput

Suggest changes →