New AWS generative AI innovations
At re:Invent 2024, we launched new products, capabilities, and features that make it easier for you to adopt generative AI at scale.
Introducing Amazon Nova
Amazon Nova is a family of new foundation models (FMs) with frontier intelligence and industry-leading price performance:
- Amazon Nova Micro, a text-only model that delivers the lowest latency responses at very low costs
- Amazon Nova Lite, a very low-cost multimodal model that quickly processes image, video, and text inputs
- Amazon Nova Pro, a highly capable multimodal model with the best combination of accuracy, speed, and cost
- Amazon Nova Canvas, a state-of-the-art image generation model
- Amazon Nova Reel, a state-of-the-art video generation model
This image was generated using Amazon Nova Canvas with the prompt "a portrait of a happy corgi dog”.
Transform how your organization works
Amazon Q is the generative AI-powered assistant from AWS that helps you streamline processes, enhance decision making, and boost productivity.
Amazon Q has many new capabilities:
- Build and deploy faster with Amazon Q Developer agents
- Transform your large-scale migration and modernization projects
- Access your Amazon Q assistant in third-party applications with simple-to-deploy extensions
- Unite your structured and unstructured data sources
- Deliver conversational self-service support for your customers
Easily build and scale generative AI applications
Amazon Bedrock is a fully managed service that helps you evaluate and choose between high-performing FMs from leading AI companies. You can build generative AI applications without the need to manage underlying infrastructure.
Amazon Bedrock has many new capabilities:
- Access to over 100 self-managed models on Amazon Bedrock Marketplace
- Specialized models for video generation
- Model distillation for cost-effective, task-specific inference
- Prompt routing and caching for cost-optimized generative AI applications
- Multi-agent collaboration for enhanced problem-solving
- Application development with Bedrock IDE in Amazon SageMaker Unified Studio
Infrastructure to build, train, and customize AI models
Amazon Elastic Compute Cloud (Amazon EC2) Trn2 instances are the most powerful Amazon EC2 instances for generative AI training and inference. Powered by AWS Trainium2 chips, Trn2 instances offer up to 30-40% better price performance the current GPU-based EC2 instances (P5e and P5en).
Three new Amazon SageMaker HyperPod innovations make it easier for you to start building your own FMs for training some of today’s most popular models – saving you up to 40% of training time. You now have full visibility and control over compute resource allocation with task governance and flexible training plans efficiently use compute resources to accelerate model training time.
With Amazon SageMaker AI, you can now discover, deploy, and use fully managed generative AI and machine learning (ML) development applications from AWS Partners in Amazon SageMaker Studio. This integration helps you choose the tools that work best for you.
Advancing trust in AI
We have a long-standing commitment to responsibly build and use AI to foster trust and deliver long-term value to our customers. We prioritize responsible AI innovation and recently announced ISO/IEC 42001 accredited certification for our core AI services.
Here are some new responsible AI capabilities:
- Build trustworthy applications for high-stakes use cases with Automated Reasoning checks in Amazon Bedrock Guardrails
- Evaluate FMs and RAG applications against key safety and accuracy criteria by using the large language model (LLM)-as-a-judge capability
- Access new AI Service Cards for Amazon foundation models
Explore our solutions
Explore our products
Training and certification
Powering customer innovation
Discover how customers across industries boost employee productivity, deliver exceptional customer experiences, and streamline business processes.