Using Claude 4.6 on Vertex AI | Model IDs for Sonnet & Opus

Can You Use Claude 4.6 on Vertex AI? — The Bottom Line

The short answer is yes: Claude 4.6 models are available on Vertex AI as a Model-as-a-Service offering. The 4.6 generation available on Vertex AI includes two models — the mid-tier Claude Sonnet 4.6 and the flagship Claude Opus 4.6. Both run on Vertex AI's fully managed infrastructure, so there is no GPU or server management required. Simply enable them from Model Garden and you can start making API calls immediately (source: Claude on Vertex AI (Anthropic Docs)).

A key feature of the 4.6 generation is that both Sonnet and Opus support a 1M token context window on Vertex. While the previous-generation Sonnet 4.5 supported 200K tokens, Sonnet 4.6 handles up to 1M tokens on Vertex.

The Two Claude 4.6 Models — Differences Between Sonnet 4.6 and Opus 4.6

"Claude 4.6" actually refers to two models with distinct characteristics on Vertex. Choose based on your use case.

• Claude Sonnet 4.6 (claude-sonnet-4-6): A mid-tier model balancing performance, cost, and speed. It is the first choice for a wide range of everyday tasks — coding assistance, summarization, classification, chat, and more. Ideal when processing large volumes of requests at low cost.
• Claude Opus 4.6 (claude-opus-4-6): Anthropic's flagship model, positioned as "the most powerful model." Best suited for demanding tasks requiring high accuracy and deep reasoning — complex coding, long-running autonomous multi-step workflows, financial analysis, and document generation.

The selection guideline is simple: start with Sonnet 4.6 and upgrade to Opus 4.6 only for tasks where accuracy or reasoning depth falls short. Since both share the same API request format, switching only requires changing the model value in the code shown below.

Claude 4.6 Model IDs on Vertex AI (Watch Out for Mix-ups)

On Vertex AI, the model is embedded in the endpoint URL rather than the request body, so knowing the exact model ID string is critical. The 4.6 model IDs listed in the official Anthropic documentation are as follows.

One important thing to note: the 4.6 models use an alias format without date suffixes (like @20250929). The previous-generation Sonnet 4.5 used a suffix format such as claude-sonnet-4-5@20250929, but Sonnet 4.6 and later use a simpler form without suffixes. Appending an unnecessary date like claude-sonnet-4-6@... will result in an unrecognized model ID error. If you get a "not found" error for 4.6 on Vertex, check for this suffix issue first (source: Claude on Vertex AI (Anthropic Docs)).

Also note that other platforms such as Amazon Bedrock use different model ID naming conventions, so do not reuse Bedrock model ID strings directly on Vertex.

What Makes Opus 4.6 Powerful? — Google Cloud's Official Positioning

Claude Opus 4.6 became generally available (GA) on Vertex AI in February 2026. Google Cloud's official blog describes Opus 4.6 as "Anthropic's most powerful model yet" and states that it "outperforms Opus 4.5 across all benchmarks" (source: Expanding Vertex AI with Claude Opus 4.6 (Google Cloud Blog)).

The blog highlights the following key strengths of Opus 4.6:

1. Complex coding and advanced automation — Strong at multi-step software development tasks.
2. Enterprise workflows — Generates documents, spreadsheets, and presentations with professional-grade quality.
3. Financial analysis — Draws insights across regulatory filings, market reports, and internal data.
4. Multi-step workflow automation — Reliably executes processes spanning dozens of tools, including error recovery.
5. Computer use — Handles workflows that depend on visual interpretation of screens and multi-step screen interactions.

In practice, use Opus 4.6 for high-stakes tasks where accuracy is paramount, and Sonnet 4.6 for high-volume processing where cost and speed matter.

How to Enable Claude 4.6 in Model Garden

If you are using Claude 4.6 on Vertex AI for the first time, start by enabling the model in Model Garden. Follow these steps:

1. Open the Vertex AI Model Garden in the Google Cloud Console with your target project selected.
2. Search for "Claude" and open the model card for the model you want to use (Claude Sonnet 4.6 or Claude Opus 4.6).
3. Click "Enable" on the model card and follow the on-screen instructions.
4. If calling from a local machine, run gcloud auth application-default login in your terminal to authenticate with GCP.
5. From there, specify the model ID, project ID, and region to make API calls.

In addition to enabling models from Model Garden cards, you can also procure them via Google Cloud Marketplace (source: Expanding Vertex AI with Claude Opus 4.6 (Google Cloud Blog)).

Calling Claude 4.6 with the AnthropicVertex SDK

The Claude API on Vertex AI is almost identical to the standard Messages API, with one difference: you pass anthropic_version as vertex-2023-10-16 in the request body. Using the official Anthropic client SDK handles this difference for you.

For Python, install the package and call the API as follows. Swap the model value between claude-sonnet-4-6 and claude-opus-4-6 to switch between the two 4.6 models.

pip install -U google-cloud-aiplatform "anthropic[vertex]"

from anthropic import AnthropicVertex

client = AnthropicVertex(project_id="MY_PROJECT_ID", region="global")

message = client.messages.create(
    model="claude-sonnet-4-6",  # Use "claude-opus-4-6" for Opus
    max_tokens=100,
    messages=[{"role": "user", "content": "Hey Claude!"}],
)
print(message)

For TypeScript, use the dedicated package:

npm install @anthropic-ai/vertex-sdk

import { AnthropicVertex } from "@anthropic-ai/vertex-sdk";

const client = new AnthropicVertex({ projectId: "MY_PROJECT_ID", region: "global" });

const result = await client.messages.create({
  model: "claude-opus-4-6",
  max_tokens: 100,
  messages: [{ role: "user", content: "Hey Claude!" }],
});

(Source: Claude on Vertex AI (Anthropic Docs))

The 4.6 Generation Supports 1M Tokens — Things to Know About Large Inputs

The headline feature of the 4.6 generation is that both Sonnet 4.6 and Opus 4.6 support a 1M token context window on Vertex. You can now include an entire large codebase or a massive collection of documents in a single request — tasks that required splitting with the 4.5 generation can be handled in one shot.

However, there is a catch. Vertex AI limits request payloads to 30MB, so sending large numbers of images or very long documents may hit this 30MB limit before reaching the token limit. When building large inputs with the expectation of 1M tokens, keep payload size in mind as well (source: Claude on Vertex AI (Anthropic Docs)).

Supported features include prompt caching, extended thinking, tool use, web search, citations, and structured output. On the other hand, the Files API input source specification, Message Batches and some other API endpoints, and Managed Agents are not supported via Vertex.

Pricing Differences: Global, Multi-Region, and Regional Endpoints

Vertex AI offers three types of endpoints, selected using the region parameter.

1. Global endpoint (recommended): Specify region="global". Dynamically routes to regions with available capacity, offering the highest availability with no pricing premium.
2. Multi-region endpoint: Specify region="us" or region="eu". Provides load balancing while keeping data resident within the specified geography. Priced at 10% above global.
3. Regional endpoint: Specify a specific region such as region="us-east1". Required for single-region data residency or provisioned throughput (dedicated capacity). Also carries a 10% premium.

This 10% premium pricing applies to newer models from Sonnet 4.5 onward, so it covers the 4.6 models — Sonnet 4.6 and Opus 4.6. If data residency requirements are flexible, the global endpoint is the first choice as it offers the highest availability without additional cost (source: Claude on Vertex AI (Anthropic Docs)).

Migrating from Claude 4.5 to 4.6 — Key Points

When migrating from the 4.5 generation to 4.6, keep the following differences in mind.

1. Model ID format changes: The 4.5 format claude-sonnet-4-5@20250929 becomes the suffix-free claude-sonnet-4-6 in 4.6. Make sure to update the string embedded in your endpoint URL.
2. Context expands from 200K to 1M tokens: Sonnet 4.6 enables large-volume inputs, but the 30MB payload limit mentioned above still applies.
3. Check your versioning policy for production: Since the 4.6 models use alias format, behavior may change when Anthropic releases updates. To minimize regression risk, either increase test coverage or regularly check availability in Vertex AI Model Garden.

Availability and supported regions are subject to change. For the latest information, search for "Claude" in Vertex AI Model Garden or check Google Cloud's Anthropic Claude model list.

Summary

Here are the key points for using Claude 4.6 on Vertex AI. The 4.6 generation includes Sonnet 4.6 (claude-sonnet-4-6) and Opus 4.6 (claude-opus-4-6), both in alias format without date suffixes and with 1M token support. Choose Sonnet 4.6 for a balanced approach, or Opus 4.6 — which "outperforms Opus 4.5 across all benchmarks" — for accuracy-critical demanding tasks. To implement, enable the model in Model Garden, authenticate with gcloud auth application-default login, then pass the model ID and region="global" to the AnthropicVertex SDK. As long as you watch out for the two key gotchas — the suffix issue in model IDs and the 30MB payload limit — migrating from 4.5 should go smoothly.