In July, we announced the addition of Meta’s Llama 3.1 models to Vertex AI Model Garden, and since then, developers and enterprises have eagerly embraced them. Today, we’re excited to introduce Llama 3.2, Meta’s next generation of multimodal models, now available on Vertex AI Model Garden.
Llama 3.2 brings new capabilities to edge devices, designed for privacy-focused, personalized AI applications. Here’s what’s new:
Llama Goes Multimodal
The 11B and 90B vision large language models (LLMs) in Llama 3.2 can analyze high-resolution images, such as charts, graphs, and image captions. These models open new opportunities in image-based search, content generation, interactive educational tools, and more.
Llama Goes Lightweight
Llama 3.2 also introduces 1B and 3B lightweight models for seamless integration with mobile and edge devices. These models enable private, low-latency AI experiences, such as on-device multilingual summarization and local AI agents, while preserving user privacy.
Key Features
Accessibility and Efficiency: The models are optimized for performance on edge devices, making AI accessible and efficient.
Privacy-Focused: Designed with responsible innovation, these models prioritize privacy and safety at the system level.
Vertex AI: A Unified Platform
Vertex AI makes it easy to experiment, customize, and deploy models like Llama 3.2. With over 160 enterprise-ready models in Model Garden, including Llama 3.2, users can select the best options to fit their specific needs.
Model-as-a-Service (MaaS) Offering
Llama 3.2’s 90B model is available in preview via our MaaS offering, with general availability coming soon. The 11B vision model will follow, allowing users to tailor and deploy these models with fully managed infrastructure and pay-as-you-go billing.
Self-Service Deployment
All four Llama 3.2 models are available for self-service deployment through Vertex AI Model Garden starting today, empowering enterprises to build AI solutions that meet their unique needs.
How Customers Are Building with Llama on Google Cloud
Shopify: Using Llama 3.1 on Vertex AI, Shopify has improved its data generation processes, providing businesses with data-driven insights more efficiently. Mike Tamir, Distinguished ML Engineer at Shopify, emphasized how Vertex AI’s infrastructure ensures dependable outputs for critical applications.
TransCrypts: The financial guide Castello has scaled rapidly, thanks to Llama 3.1 on Vertex AI. Zain Zaidi, Co-Founder & CEO of TransCrypts, highlighted the speed and cost-efficiency of deploying advanced models using Google Cloud’s infrastructure.
BMC: By integrating Llama 3.1 into the BMC Helix platform, BMC has enhanced conversational AI, recommendations, and IT service management. Margaret Lee, GM and SVP at BMC, noted the improvement in accuracy and the impact on customers’ access to advanced AI solutions.
Why Use Llama 3.2 on Google Cloud?
Experiment with Confidence: Vertex AI offers simple API calls and a comprehensive evaluation service for exploring Llama 3.2 capabilities in an intuitive environment.
Customizable AI: Fine-tune Llama 3.2 using your own data to create tailored solutions for your business.
Ground AI in Reality: Ensure trustworthy AI outputs by grounding models in enterprise systems and leveraging Vertex AI’s retrieval-augmented generation (RAG) features.
Intelligent Agents: Use tools like LangChain on Vertex AI to build intelligent agents powered by Llama 3.2.
Streamlined Deployment: With flexible auto-scaling, pay-as-you-go pricing, and robust infrastructure, scaling Llama 3.2 applications has never been easier.
Enterprise-Grade Security: Deploy with confidence using Meta’s Llama Guard and Google Cloud’s built-in security, privacy, and compliance features.
Start Building with Llama 3.2
Our collaboration with Meta underscores a commitment to delivering world-class AI innovation. As we continue to expand the AI ecosystem with partners like Meta, Llama 3.2 is another step in empowering enterprises with cutting-edge AI tools. Start exploring the possibilities with Llama 3.2 on Google Cloud today!
