Using via Cloud: Bedrock / Vertex

Why Choose Cloud-Via

You can call Claude or GPT directly via API, so why choose via AWS Bedrock or Google Vertex AI? The reason is ease of embedding into enterprise systems.

Cloud-Via Merits

1. Security/compliance: data stays within AWS/GCP. Integrated into internal rules via VPC/IAM
2. Billing integration: rolls into existing AWS/GCP billing. Easier for accounting too
3. Geographic constraints: can specify data location country (cases requiring in-EU processing)
4. SLA / support: if you have an enterprise contract, the responsibility boundary rides on the existing contract
5. Multi-model: try multiple models like Claude, Llama, Titan, Mistral with the same API

3 Main Platforms

	AWS Bedrock	Google Vertex AI	Azure AI Foundry
Claude	○	○	—
Gemini	—	◎	—
GPT	—	—	◎
Llama	○	○	○
Pricing	Usage + Provisioned Throughput	Usage	Usage + PTU

Calling Claude on AWS Bedrock

1. Request Model Access

AWS Console → Bedrock → Model access → check Claude models → send request (approved in minutes to hours)

2. Set IAM Permissions

Grant the IAM user bedrock:InvokeModel permission

3. Code Example (Node.js)

import { BedrockRuntimeClient, InvokeModelCommand }
  from "@aws-sdk/client-bedrock-runtime";

const client = new BedrockRuntimeClient({ region: "us-east-1" });

const body = JSON.stringify({
  anthropic_version: "bedrock-2023-05-31",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Things to see in Tokyo?" }],
});

const response = await client.send(new InvokeModelCommand({
  modelId: "anthropic.claude-opus-4-5-v1:0",
  body,
}));

const json = JSON.parse(new TextDecoder().decode(response.body));
console.log(json.content[0].text);

Calling Gemini on Google Vertex AI

1. Project Prep

GCP → Vertex AI → Enable

2. Authentication

gcloud auth application-default login for local auth

3. Code Example

import { VertexAI } from "@google-cloud/vertexai";

const vertex = new VertexAI({
  project: "your-project",
  location: "us-central1",
});

const model = vertex.getGenerativeModel({
  model: "gemini-2.5-pro",
});

const result = await model.generateContent("Things to see in Tokyo?");
console.log(result.response.text());

Abstract with Vercel AI SDK

The Vercel AI Gateway is handy for cross-cloud code unification:

import { generateText } from "ai";
import { gateway } from "@ai-sdk/gateway";

// Specify by string without minding provider/base differences
await generateText({
  model: "anthropic/claude-opus-4-5",  // uses Bedrock-via behind the scenes
  prompt: "...",
});

How to Choose

• Already on AWS & want Claude → Bedrock
• On GCP & Gemini-centric → Vertex AI
• Azure & GPT-centric → Azure AI Foundry
• Startup/personal dev → Anthropic / OpenAI directly suffices

Cautions

• Feature gap: Bedrock's Claude may lag the direct API in features (Computer Use, etc.)
• Price difference: cloud-via adds a bit. In exchange, fixed pricing via Provisioned Throughput
• Region restriction: regions where models are available are limited