Anthropic’s Claude 3 Models in Google Cloud Vertex AI

Claude 3 in Vertex AI matters less as a model announcement than as a platform decision. For teams already standardizing on Google Cloud, the key architectural change is that a third-party frontier model becomes available through the same managed interface as other Vertex AI options. That reduces the amount of custom glue code, procurement friction, and environment sprawl that often accompanies multi-model AI programs.

The practical trade-off is choice versus governance. Having Opus, Sonnet, and Haiku in one catalog lets architects map capability to workload: deeper reasoning for complex tasks, faster responses for latency-sensitive flows, and a more compact option where cost and throughput matter. But multi-model availability also increases the burden on platform teams to define routing rules, approval processes, prompt/version controls, and evaluation criteria so usage stays consistent across applications.

Managed APIs also shift operational responsibility. Teams can avoid standing up dedicated inference infrastructure, but they still need to plan for integration with existing identity, network, observability, and data-handling controls. In practice, the adoption question becomes less about whether the model can be called, and more about how well it fits enterprise guardrails, workload isolation, and lifecycle management inside an existing cloud operating model.

The broader implication is that AI architecture is becoming more composable. As model catalogs expand, engineering organizations will increasingly select providers and model sizes per use case rather than committing to a single default. That can improve agility, but only if platform engineering, security, and application teams align on shared standards for deployment, cost oversight, and responsible use.


At Google Cloud, we’re committed to empowering customer choice and innovation through our curated collection of first-party, open-source, and third-party models available in Vertex AI. That’s why we’re thrilled to announce that Claude 3 — Anthropic’s new family of state-of-the-art models — will be generally available in Vertex AI Model Garden over the coming weeks, including private preview access for one of the models starting today.

Claude is Anthropic’s next-generation AI assistant that helps manage organizations’ tasks, no matter the scale. Anthropic’s launch of Claude 3 includes a family of three distinct models optimized for various enterprise applications:

  • Claude 3 Opus: Anthropic’s most capable and intelligent model yet.
  • Claude 3 Sonnet: Anthropic’s best combination of skills and speed.
  • Claude 3 Haiku: Anthropic’s fastest, most compact model.

Compared to earlier iterations of Claude, both Claude 3 Opus and Sonnet offer superior reasoning across complex tasks, content creation, scientific queries, math, and coding, while Haiku is Anthropic’s fastest and most cost-effective model. All Claude 3 models boast improved fluency in non-English languages, as well as vision capabilities that unlock tasks ranging from image metadata generation to insights extraction across PDFs, flow charts, and a diverse range of other formats.

In the weeks ahead, Google Cloud customers will be able to select from all three Claude 3 models via API access in Vertex AI Model Garden. And starting today, customers can apply for private preview access to Claude 3 Sonnet in Model Garden.

Build and deploy with Claude 3 in Vertex AI

Through our partnership, we will bring Anthropic’s latest models to our customers via Vertex AI, the comprehensive AI development platform. The Claude 3 family joins over 130 models already available in Vertex AI Model Garden, further expanding customer choice and flexibility as gen AI use cases continue to rapidly evolve.

By making Claude 3 models available in Vertex AI, customers have powerful new options to:

  • Accelerate AI development with quick access to Claude’s pre-trained models through simple API calls in Vertex AI.
  • Focus on applications, not infrastructure as Claude models are offered in Vertex AI as managed APIs — meaning customers can concentrate on building groundbreaking applications instead of worrying about backend complexity or the management overhead of underlying infrastructure.
  • Optimize performance and costs by leveraging flexible auto-scaling and pay-only-for-what-you-use pricing to optimize costs as needs grow. And of course, leverage world-class infrastructure, purpose-built for AI workloads.
  • Deploy responsibly with Google Cloud’s built-in security, privacy, and compliance as Vertex AI’s assortment of models and tools are offered with Google Cloud’s enterprise-grade security, privacy, and compliance for generative AI.

Sign up to access Claude 3 in Vertex AI

This is just the beginning of our partnership with Anthropic, and we’re excited to enable customer innovation with the newest models. We’ll continue to work closely with Anthropic and other partners to keep our customers at the forefront of AI capabilities.

To get started with Vertex AI, visit our product page. To access Claude 3 Sonnet via private preview, visit Model Garden, and to learn more about Claude 3, check out Anthropic’s announcement.

Original Post

Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.