Where Each Model Is Actually Available

"Available" isn't one fact about a model — it's available somewhere, at some price, under some set of features, and all three of those can diverge sharply depending on which platform is actually serving the request.

Direct API is the baseline this site prices

Every figure elsewhere on this site — pricing, rate limits, service tiers — is sourced from OpenAI's own direct API documentation. That's the baseline every other platform's numbers should be checked against, not assumed to match.

Third-party hosting platforms diverge in real, specific ways

Amazon Bedrock bills through AWS on its own schedule, explicitly stated to differ from OpenAI-direct pricing — this site has no basis to publish what that AWS-side number actually is, and doesn't try to. Google Vertex AI and Microsoft Foundry carry their own separate availability and pricing that likewise isn't sourced here. And the Batch API specifically is not offered on any of these three third-party platforms at all — a capability gap, not just a pricing difference, worth checking before designing a cost strategy around Batch for a deployment that doesn't run through OpenAI directly.

Why this matters for planning, not just billing

A workload's real platform choice is rarely made on pricing alone — existing cloud infrastructure, procurement relationships, and compliance requirements all pull toward one platform or another. What this page exists to prevent is discovering a platform-specific gap — a missing capability, a different billing schedule — after that architectural decision is already locked in, rather than as part of making it.

What to check before committing to a platform

If Batch pricing matters to your cost model, confirm you're calling OpenAI directly. If you're on Bedrock, Vertex or Foundry, get pricing from that platform's own documentation, not from OpenAI's pricing page — this site's own numbers simply don't apply there.