Claude is the family of AI models built by Anthropic. When people say "Claude 3P," they are not naming a model. They are describing where you buy and run it. "3P" is shorthand for "third-party platforms" — accessing Claude through a cloud provider you already use, rather than directly from Anthropic. The models are the same; the front door is different.
1P and 3P in one paragraph each
1P (first-party) means going straight to the source: the Claude API from Anthropic at anthropic.com. You sign up with Anthropic, get an API key, and Anthropic bills you. This is the simplest path if you have no strong cloud attachment and want every feature the moment it ships.
3P (third-party platforms) means reaching the same Claude models through a cloud provider's marketplace and infrastructure. Your existing cloud account handles access control, billing, and much of the governance. For an enterprise that already has a cloud contract, security review, and procurement process in place, this can shave months off the path from "interesting demo" to "approved production system."
The four third-party platforms
As of July 2026 there are four 3P routes to Claude. Three are operated by the cloud providers themselves; one is operated by Anthropic but lives inside AWS.
| Platform | Operated by | Runs on |
|---|---|---|
| Claude Platform on AWS | Anthropic | AWS |
| Amazon Bedrock | AWS | AWS |
| Google Vertex AI | Google Cloud | Google Cloud |
| Microsoft Foundry | Microsoft | Azure |
The distinction in the "operated by" column matters more than it looks. Bedrock, Vertex AI, and Foundry are managed services run by the cloud providers, so new Claude features arrive on each provider's own schedule. Claude Platform on AWS is Anthropic-operated and gets same-day parity with the first-party API — new capabilities typically land the day they ship — while still running inside AWS.
Why 3P exists at all
Enterprises adopt 3P for reasons that have little to do with the models themselves.
Procurement is already done. Buying through your cloud marketplace usually means the spend flows through an existing agreement, often counting toward committed cloud spend. No new vendor onboarding, no new master service agreement to negotiate from scratch.
Security and identity plug in. Access to Claude is governed by the same identity and permission systems your teams already use — IAM roles on AWS, service accounts on Google Cloud, resource keys on Azure. Your audit and monitoring tooling sees Claude traffic the way it sees any other cloud service.
Compliance posture is inherited. Running Claude inside your existing cloud boundary means it largely inherits that cloud's compliance posture — though you should confirm specifics for your industry and region with your provider rather than assuming.
What's the catch?
Feature timing and coverage. The core capabilities — the Messages API, streaming, tool use, vision, adaptive thinking, and prompt caching — are available on all four platforms. But some newer capabilities lag: the Batch API, Files API, code execution tool, and web fetch tool are not available on Bedrock or Vertex AI, and some of Foundry's broad feature set is still in beta. List prices match the first-party API across the cloud marketplaces; discounts for committed use are negotiated with your cloud provider, not Anthropic.
The practical takeaway: 3P is rarely a question of "is it good enough?" and usually a question of "which door matches the house we already live in?" If your organization runs on one cloud, the default answer is the Claude offering on that cloud — with Claude Platform on AWS as the notable option when you want Anthropic-operated, first-party feature velocity without leaving AWS.
Where to go next
For the trade-offs in more depth, read 1P vs 3P: Two Ways to Buy Claude, then use the platform decision framework to pick your door. The quickstart gets you from zero to a first API call.