Amazon Bedrock

Amazon Bedrock is a service from Amazon Web Services for building and scaling generative AI applications, offering users access to foundational models (FMs) from AI start-ups and Amazon. First announced on April 13, 2023, Bedrock's FMs are hosted on AWS, allowing access through an API without users having to host the large amount of data required on their own infrastructure. Additionally, Bedrock's enterprise clients can quickly customize FMs with their own data, integrating and deploying applications using other AWS tools. Data provided into Bedrock's FMs is not used to train future iterations of Amazon AI models.

FMs perform a wide range of tasks. However, they can be customized for domain-specific functions with additional training that requires only a small fraction of the data and computational resources needed to train a model from scratch. Bedrock aims to help enterprises build their own generative AI applications and create unique customer experiences specific to their company's operations, voice, style, and services by offering access to a range of FMs, each with different strengths and characteristics.

Upon release, Bedrock is in limited preview and is not available to new customers. Amazon also did not announce pricing plans for the service.

Bedrock provides access to multiple FMs, allowing customers to select the best model for their needs. This includes models from third-party AI start-ups, such as AI21 Labs, Anthropic, and Stability AI, as well as Amazon's own Titan family of FMs.

• Jurassic-2 (AI21 Labs)—multilingual large language models (LLMs) that can generate text in Spanish, French, German, Portuguese, Italian, and Dutch
• Claude (Anthropic)—a model capable of a range of conversational and text-processing tasks
• Stable Diffusion (Stability AI)—a text-to-image model able to generate images, art, logos, and graphic designs
• Titan (Amazon)—a text-generating model for performing tasks such as text-generation, summarizing documents, and extracting information from databases, and an embedding model that translates text inputs into numerical representations