Cohere Command Light and Embed Models Now Available on Amazon Bedrock
Cohere is pleased to announce the availability of Cohere Command Light and Cohere Embed English and multilingual models on Amazon Bedrock. These models join the existing Cohere Command model.
Amazon Bedrock is a fully managed service that offers a range of high-performing foundation models (FMs) from leading AI companies, including Cohere, AI21 Labs, Anthropic, Meta, Stability AI, and Amazon. It provides a wide range of capabilities to simplify the development of generative AI applications while ensuring privacy and security. This launch expands the model choices available on Amazon Bedrock, enabling you to build and scale enterprise-ready generative AI applications. You can find more information about Amazon Bedrock in Antje’s post.
Cohere’s flagship text generation model is Command. It is designed to follow user commands and be useful in business applications. Embed, on the other hand, is a set of models trained to generate high-quality embeddings from text documents.
Embeddings are a fascinating concept in machine learning (ML) and are central to many applications that process natural language, recommendations, and search algorithms. Embeddings allow you to transform any type of document, such as text, image, video, or sound, into a suite of numbers known as a vector. These vectors capture meaningful information, semantic relationships, and contextual characteristics. In simple terms, similar documents are represented by vectors that are “close” to each other, capturing semantic similarity as perceived by humans. Cohere Embed is a family of models trained to generate embeddings from text documents. It is available in both English and multilingual versions on Amazon Bedrock.
Text embeddings have three main use cases:
- Semantic searches: Embeddings enable searching collections of documents based on meaning, leading to search systems that better incorporate context and user intent compared to traditional keyword-matching systems.
- Text classification: Build systems that automatically categorize text and take action based on the type. For example, an email filtering system might route one message to sales and escalate another message to tier-two support.
- Retrieval Augmented Generation (RAG): Improve the quality of a large language model (LLM) text generation by augmenting prompts with additional data provided in context. This data can come from various sources, such as document repositories, databases, or APIs.
To illustrate the use of Cohere Embed, imagine you have hundreds of documents describing your company policies. Due to the limited size of prompts accepted by LLMs, you need to select relevant parts of these documents to include as context in prompts. To solve this, you can transform all your documents into embeddings and store them in a vector database, such as OpenSearch. When a user wants to query this corpus of documents, their natural language query is transformed into a vector and a similarity search is performed on the vector database to find the most relevant documents. The original query and the relevant documents are then combined in a prompt for the LLM, improving the accuracy and relevance of the generated answers.
You can now integrate Cohere Command Light and Embed models into your applications written in any programming language by calling the Bedrock API or using the AWS SDKs or AWS Command Line Interface (AWS CLI).
For more information and code examples, please refer to the Amazon Bedrock workshop.
The Cohere Embed models are available to all AWS customers in the US East (N. Virginia) and US West (Oregon) regions. AWS charges for model inference, with different pricing structures for Command Light and Embed models. For more details, please visit the Amazon Bedrock pricing page.
Start leveraging the power of text embeddings with Amazon Bedrock and the Cohere Embed models in your applications today. Happy building!
— seb
