October 16, 2026 · Ahmed · More by Ahmed
OpenAI's Build, Launch, Grow: What the New API Tiers Gate
OpenAI collapsed five usage tiers into three — Build, Launch, and Grow. Here's what each tier's threshold and rate limits mean for pipeline throughput.
On October 6, 2026, OpenAI's API changelog announced it had simplified usage tiers from five down to three: Build, Launch, and Grow. For teams running automation pipelines against the OpenAI API — bulk processing jobs, CRM-sync-style outreach automation, agentic workflows — this isn't just a cosmetic rename. It changes the thresholds that gate how much throughput you can push, and it's worth knowing exactly where your pipeline's ceiling now sits.
What Each Tier Requires and Caps
Organizations upgrade automatically as total credit purchases cross each tier's threshold — no manual application. Per OpenAI's rate-limits guide:
- Build: $5 in total credit purchases, $500/month usage limit.
- Launch: $100 in total credit purchases, $5,000/month usage limit.
- Grow: $500 in total credit purchases, $200,000/month usage limit.
OpenAI's Help Center confirms existing paid organizations were migrated automatically — old Tiers 1-2 moved to Build, Tiers 3-4 to Launch, and Tier 5 to Grow — with no additional payment required, and API pricing and any custom rate limits left unchanged by the migration.
Rate Limits Scale by Tier and Model Family
Standard requests-per-minute (RPM) and tokens-per-minute (TPM) ceilings scale with tier. For the Astra, Sol, and Terra model family: Build gives 5,000 RPM / 1,000,000 TPM, Launch gives 10,000 RPM / 4,000,000 TPM, and Grow gives 15,000 RPM / 40,000,000 TPM. The Luna family gets a separately higher allowance at every tier: 5,000 RPM / 2,000,000 TPM at Build, 10,000 RPM / 10,000,000 TPM at Launch, and 30,000 RPM / 180,000,000 TPM at Grow. If your pipeline is batching large volumes through a bulk-processing or CRM-sync job, these are the numbers that determine how fast you can safely ramp request volume as you scale.
Check Which Service Tier You're Actually On
OpenAI's guide is explicit that these Standard rate limits differ from Ultrafast-tier rate limits, its separate low-latency service tier. A pipeline evaluating its throughput headroom needs to confirm which service tier its requests actually run under before reading the Standard limits as the ceiling — the two aren't interchangeable, and the Standard numbers above don't describe Ultrafast capacity.