Anthropic launches Google Cloud service with same-day Claude feature access

Google Cloud customers can now get Claude’s full developer platform without moving authentication and billing outside Google Cloud. Anthropic launched Claude Models on Google Cloud (Anthropic managed) on October 22, 2026, promising new features and betas on their native release day. The distinction is who runs it: Anthropic operates the service and processes the data.

Google handles access and the invoice

The offering works with existing Google Cloud Identity and Access Management, or IAM, credentials. Teams can use the permissions they already manage rather than adopt another authentication system. Claude charges appear on a single Google Cloud invoice, and usage counts against an existing Google Cloud spending commitment.

That purchasing arrangement does not make Google the service operator. Anthropic is the data processor for this route. A separate Google-managed Claude offering remains available, with Google as the data processor. Customers therefore have two options with different operating responsibilities, not simply two ways to pay for the same service.

Native tools, including the betas

Anthropic says the new offering includes the complete Claude Platform feature set, with future features and betas arriving the same day as on its native platform. Available models are Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. New models are also due to arrive as they launch.

• Agent work: Managed Agents, in beta, supplies infrastructure for long-running agents. Code execution runs Python and analyzes data within a developer request. Computer use lets Claude operate a browser or desktop, while agent skills provide repeatable workflows.
• Documents and outside information: The Files API supports uploading and referencing documents across conversations. Citations ground responses in source documents; web search and web fetch bring in current online information. The beta MCP connector connects Claude to other tools without client code.
• Cost and workload controls: Prompt caching reduces costs and response delays when context is repeated. Batch processing handles high-volume work that does not require an immediate response. The beta Advisor strategy lets an agent consult an advisor model.

Fast Mode for Opus 5.5 is generally available through this offering. Anthropic says it generates output tokens up to 2.5 times faster, positioning it for live debugging and interactive agents. Listed rates are $8 per million input tokens and $40 per million output tokens.

Customers also receive the Claude Console, Anthropic’s development environment for managing agents. It provides tools to write prompts, configure agents and run them in hosted environments. Teams can monitor usage and costs there, and manage access and security for each workspace.

Where the Google-managed option still fits

The Anthropic-managed service currently offers US and Global endpoints. Anthropic recommends the Google-managed option for customers who need Google to process their data, have strict requirements about the regions where data stays, or want native Google Cloud integrations such as Model Armor or Agent Engine.

There is also a purchasing step for customers with negotiated discounts. Anthropic’s launch announcement tells them to contact their Anthropic or Google Cloud account executive before starting, so discounts are applied correctly. Discounts cannot be applied retroactively to usage—a concrete reason to settle account terms before sending requests through the new offering.