Vertex AI Anthropic Claude Pricing Details: Tokens, Cache, and Billing

If you searched for “cloud.google.com vertex ai anthropic claude pricing details,” start with Google's official pricing page and select the exact Claude model, location, and billing mode. Claude API costs are usage charges rather than a Claude app subscription: compare input, output, and any cache-related line items separately before estimating a monthly bill.

Where are the official Claude prices on Google Cloud?#

Use the Google Cloud generative AI pricing page as the authority for your cloud deployment. That URL redirected to Google's Agent Platform pricing page when checked on 2026-10-15. Navigation and product labels can change, so verify the destination and locate the partner-model section instead of assuming an old screenshot is current.

Choose the model version first. A price for one Opus generation does not establish a price for another generation, and Sonnet and Haiku are separate model families. Then check the selected location, supported access mode, and any terms attached to the displayed rate. If you cannot find an applicable rate, obtain it from the cloud console or sales rather than extrapolating.

What does a Claude token price include?#

Input tokens represent the content sent to the model; output tokens represent the generated answer. Count the complete request, including instructions and previous conversation content, rather than just the latest question. Longer answers can dominate a bill even when input is inexpensive.

Cache reads and cache writes can have different rules from ordinary input. A repeated prompt is not proof of a billable cache hit: use the usage fields and billing records. Likewise, an asynchronous batch offer should not be applied to a real-time request unless the relevant platform explicitly supports it for that model.

For uncached usage, estimate cost as input tokens multiplied by the input rate, plus output tokens multiplied by the output rate, with each token count divided by one million. Add other documented line items separately. This estimate does not include taxes, currency conversion, or unrelated cloud services.

What is a sourced October 2026 comparison?#

The Claude Opus 4.8 pricing page records the following comparison as of 2026-10. Its cloud figures are maintained reference entries, not proof that Google's current regional quote was independently verified today. The official page remains authoritative for a Vertex purchase.

Access route	Input per 1M tokens	Output per 1M tokens	Scope
Google Vertex AI reference	$5.00	$25.00	Reference entry on the linked model page
Anthropic reference	$5.00	$25.00	Provider reference on the same page
Crazyrouter	$3.25	$16.25	Published gateway schedule

At those rates, an illustrative workload of one million uncached input tokens and one million output tokens totals $30.00 at the reference rate or$19.50 at the gateway rate. This is arithmetic from the cited schedule, not a measured production workload or a contractual quote. Region, account terms, supported features, and retry behavior still need review.

How should you compare Vertex AI with a gateway?#

Keep procurement and application needs in the comparison. A Google Cloud deployment may fit an existing cloud account, access-control setup, and billing workflow. A gateway can simplify a mixed-provider application, but it introduces another service relationship and should be evaluated on data terms and model routing as well as price.

Check Crazyrouter's pricing overview for the selected model and its model catalogue for discovery. Do not reuse a Google Cloud deployment identifier as a gateway model name without checking. These routes also have different authentication and endpoint conventions; this pricing article does not present a gateway request as a working Vertex API example.

FAQ#

Is Claude on Vertex AI included in a Claude Pro subscription?#

A consumer Claude subscription and cloud API usage are separate products. Budget the cloud model charges against the project and billing agreement that actually process your requests.

Can I assume all regions have the same Claude price?#

No; check the applicable model, location, and billing mode on the official cloud page. A reference comparison table should not replace the quote for your deployment.

Is Crazyrouter always cheaper than Vertex AI?#

The cited Opus 4.8 snapshot has a lower published gateway token rate than its Vertex reference entry. That does not establish the cheapest total cost for every contract, cache pattern, or workload; compare the current applicable rates and required features before choosing.

<!-- [crazyrouter](https://docs.crazyrouter.com/en/introduction)-related-links -->

Related on crazyrouter.com#

• OpenRouter alternatives