According to a post published by Google Cloud, Claude Opus 5.5 and Claude Sonnet 5.5 are now available in Google Antigravity for Google Cloud customers. The move is made in keeping with Google Cloud's commitment to an open ecosystem and to provide developers with the right model for every task, ranging from rapid bug fixing to planning a complex migration. The Claude Opus 5.5 model is built for complex work, while Claude Sonnet 5.5 serves as a faster and lower-cost complement.
Claude Model Availability and Development Environment Support
Once enabled by an administrator, developers can select Claude Opus 5.5 and Claude Sonnet 5.5 directly from the model picker. At the same time, administrators can also select Google's Gemini models, including Gemini 3.8 Flash and Gemini 4 Argon.
The models apply across Antigravity 2.0, the Antigravity CLI, and the recently released IDE extensions for VS Code, JetBrains IntelliJ, and Visual Studio. This support allows choosing the appropriate model for the task under a unified enterprise security and billing framework. According to the post, having multiple model families available inside the same agent harness unlocks practical multi-model workflows that improve code quality while controlling latency and costs.
Pattern 1: Cross-Model Verification in Loop and Graph Workflows
The first engineering pattern presented in the post is cross-model verification ("LLM-as-a-Judge") within agentic loops or multi-step graph workflows. When building these workflows, using the same model both to write code and to review its own output can reinforce model-specific blind spots. Pairing two distinct model families—one as the implementer and another as the validator (judge)—provides independent verification before moving to the next stage of the workflow.
In Antigravity, this pattern can be operated interactively or via automated subagents:
- Plan and implement with Model A: Begin work in planning mode using Claude Opus 5.5 or Gemini 4 Argon to design and implement a multi-file change.
- Validate with Model B: Before committing or advancing the loop to the next stage, the developer switches models, or invokes a one-shot check using the /model command in the CLI, allowing a different model family to critique the implementation against the test suite and requirements.
Pattern 2: Pairing Reasoning Models with Flash Subagents
The second engineering pattern focuses on handling complex engineering tasks, which frequently combine a few decisions requiring high reasoning capacity alongside dozens of high-volume, mechanical lookups and checks. These lookups include searching files, grepping logs, inspecting lint tool output, or classifying test failures. Running every lookup and check operation on a frontier model can increase both latency and financial spend.
Antigravity's subagent customization system (configured via files under the path .agents/agents/*.md) allows assigning specific model tiers to dedicated subagents:
- inherit tier: Subagents configured with this model tier use the active session model from the main conversation for subtasks that require deep reasoning.
- flash tier: Subagents configured with this model tier offload high-speed search, classification, and triage tasks to Gemini Flash models.
The post presents two examples of subagent configuration files:
The first file, test-triager.md, defines a subagent named test-triager operating on the flash tier with the read and edit tools. Its role is to run unit tests, classify failures by root cause (compilation error, assertion failure, or timeout), and return a structured and concise summary of the failures to the primary agent.
The second file, gemini-judge.md, defines an independent verifier subagent named cross-model-judge operating on the pro tier using Gemini Pro, equipped with the read tool. Its role is to review the workspace code changes (diff) against the user's requirements, check for regressions, missing tests for edge cases, and security issues, and return a pass/fail verdict with exact line references.
Through this configuration, rapid testing and classification loops can be delegated to test-triager, while cross-model-judge can be triggered to validate the final diff, thereby combining the strengths of each model within a single workflow.
Getting Started, Account Management, and Enterprise Billing
Claude Opus 5.5 and Claude Sonnet 5.5 are now available to Google Cloud customers using Antigravity, with usage instructions outlined by role:
For administrators in a Gemini for business environment: Third-party partner models do not draw from the pooled monthly token quota of the Gemini Enterprise Standard and Gemini Enterprise Plus editions. Instead, their usage is billed on a pay-as-you-go basis through the organization's existing Google Cloud Billing account; therefore, administrators must ensure that overage billing is enabled. To configure overage billing, set spend limits, and authorize partner models for the organization, administrators navigate in the Google Cloud Console to "Gemini Enterprise", then to "Settings", and then to "AI developer tools".
For administrators in a Gemini platform environment: Administrators can also enable partner models if Antigravity usage takes place through the API in Gemini platform. To do so, they must enable the Claude models in Model Garden and complete Anthropic's standard terms of service.
For developers: Users should update to the latest version of Antigravity 2.0, the Antigravity CLI, or the IDE extensions, and open the model picker (or run the /model command) to select the desired model for the task. Developers can switch the active model between turns within the same conversation thread while preserving workspace state, open artifacts, and terminal context.
An additional note in the post emphasizes that lightweight first-party Gemini models continue to run in the background to handle frequent client-side coordination, such as context management and summary generation. These models consume first-party tokens and may incur charges.