Introducing Gemma 3 27B Instruct: Now Available on Amazon Bedrock Marketplace and SageMaker JumpStart

Accelerate Your Generative AI Solutions with a Powerful Language Model

Explore how to kickstart your journey with Gemma 3 27B Instruct and unlock its advanced features for your AI applications.

Exciting News: Gemma 3 27B Instruct Models Launch on AWS

We are thrilled to announce the availability of Gemma 3 27B Instruct models through the Amazon Bedrock Marketplace and Amazon SageMaker JumpStart. Developers and data scientists can now leverage this powerful 27-billion-parameter language model, along with its specialized instruction-following variants, to accelerate the building, experimentation, and deployment of generative AI solutions on AWS.

In this blog post, we’ll guide you on how to get started with the Gemma 3 27B Instruct model on both Amazon Bedrock Marketplace and SageMaker JumpStart, showcasing the model’s robust instruction-following capabilities for your applications.

Overview of Gemma 3 27B

Gemma 3 27B is a state-of-the-art, open-weight multimodal language model developed by Google. Designed for exceptional efficiency and contextual understanding, it processes both text and image inputs seamlessly. Key innovations include:

• Redesigned Attention Architecture: Enhances the way the model understands and generates text.
• Multilingual Support: Offers robust support for over 35 languages, with training exposure to more than 140, ensuring global usability.
• Extended Context Capabilities: Capable of handling up to 128,000 tokens, making it ideal for intricate reasoning and long-form interactions.
• Memory-efficient Inference: Incorporates architectural updates to improve performance while reducing resource usage.

Key Features:

• Multimodal Input: Unified reasoning across text, images, and short videos.
• Advanced Reasoning: Ideal for complex tasks, automated content generation, chatbots, and virtual assistants.
• Function Calling: Facilitates the building of workflows through natural-language interfaces.

Key Use Cases:

• Q&A and Summarization: Streamlining long documents into concise summaries.
• Visual Understanding: Including image captioning, object identification, and document analysis.
• Multilingual Applications: Creating AI-driven tools that cater to diverse languages.
• Automated Workflows: Using function calling to create interactions with external systems.

Deploying Gemma 3 27B Instruct on Amazon Bedrock Marketplace

Amazon Bedrock Marketplace simplifies access to a wide array of foundation models (FMs), including Gemma 3 27B Instruct.

Prerequisites:

1. An active AWS account.
2. Access to accelerated instances (GPUs) for hosting large language models (LLMs).

Steps to Deploy:

1. Go to the Amazon Bedrock console and navigate to Model catalog.
2. Filter by Gemma and select Gemma 3 27B Instruct.
3. Review model details, pricing, and implementation guidelines before choosing Deploy.
4. Set the Endpoint name and Number of instances based on your needs (1-100).
5. Choose an appropriate instance type, like ml.g5.48xlarge.
6. Monitor the deployment status in the Managed deployments section.
7. Once deployed, test model capabilities in the Amazon Bedrock playground.

For code examples to invoke the model using Bedrock APIs, you can follow a simple script provided in the original announcement.

Deploying Gemma 3 27B Instruct with SageMaker JumpStart

SageMaker JumpStart allows users to deploy pre-trained models effortlessly.

Methods of Deployment:

1. Using the SageMaker JumpStart UI: A user-friendly interface enables model selection and customization without extensive coding.
2. Programmatically with the SageMaker Python SDK: Ideal for users who prefer hands-on coding and flexibility.

Steps to Deploy via the UI:

1. Access the SageMaker Studio console and choose JumpStart in the navigation.
2. Look for Gemma 3, click the model card for more details, and hit the Deploy button.
3. Configure the Endpoint name, Instance type, and Initial instance count.
4. Review your configurations and select Deploy.

Programmatic Deployment Example:

Using the SageMaker Python SDK, you can deploy the model with a few lines of code. Install the SDK, set your permissions, and follow the provided example to get started.

Running Inference:

After deploying the model, it’s straightforward to use the SageMaker API for making predictions. The original article provides example code for this as well.

Clean Up Your Resources

To avoid ongoing charges, ensure you clean up all deployed endpoints:

1. Delete SageMaker Endpoints: Remove any endpoints associated with the Gemma 3 model.
2. Review Bedrock Resources: Ensure no extra endpoints are active.

Conclusion

The launch of Gemma 3 27B Instruct models through Amazon Bedrock Marketplace and SageMaker JumpStart enables developers, researchers, and businesses to create advanced generative AI applications with ease. Its high performance, multilingual capabilities, and the reliability of AWS infrastructure make it an excellent choice for various use cases.

We invite you to explore the Gemma 3 27B Instruct models available now on the SageMaker JumpStart console or Amazon Bedrock Marketplace. Dive into the examples and start building the next generation of generative AI solutions!

About the Authors

Santosh Vallurupalli is a Sr. Solutions Architect at AWS, focusing on networking and migrations. He enjoys traveling and watching Formula 1.

Aravind Singirikonda is an AI/ML Solutions Architect at AWS, assisting healthcare and life sciences customers with AI solutions.

Pawan Matta specializes in scalable architectures in the gaming sector, with a passion for FIFA and cricket.

Ajit Mahareddy has over 20 years of experience in product management and generative AI technologies.