Cohere in Amazon Bedrock

Build enterprise AI applications that understand your business

Get started with Cohere models in Amazon Bedrock

Introducing Cohere’s enterprise foundation models

Command R and R+ are Cohere's powerful, advanced language models for real-world enterprise applications. They balance efficiency and accuracy, enabling businesses to move from proof-of-concept to daily AI utilization. Supporting 10 key languages, these models excel at retrieval-augmented generation (RAG) and long-context tasks. Ideal for global enterprises, Command R and R+ are optimized for RAG use cases and adept at text generation. They're well-suited for full-scale AI implementation, with R+ offering enhanced performance for businesses ready to leverage AI across operations.

Cohere's Embed 3 is an industry-leading embeddings model that generates embeddings from both text and images. It enables enterprises to unlock value from vast image data, creating accurate search systems for complex reports, product catalogs, and design files. Supporting 100+ languages and excelling in multimodal search tasks, Embed 3 streamlines advanced AI applications, enhancing ecommerce experiences, design asset management, and data-driven decision-making processes.

Cohere's reranker model, Cohere Rerank 3.5, provides a powerful semantic boost to the search quality of any keyword or vector search system. In RAG use cases, reranking can help ensure that only the most relevant information is passed to the model. This results in better responses, reduced latency, and lower costs because the model processes less information.

Benefits

With a context window of up to 128K tokens, the Command R models understand and generate responses within a broad context, making them ideal for complex workflows with large document ingestion, relevant citations with advanced retrieval, and tool use.
The Command R models have the capability for multilingual generation across 10 key business languages including: English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Arabic, and Chinese.
Command R+ supports multi-step tool use which allows the model to combine multiple tools over multiple steps to accomplish difficult tasks. The model can even correct itself when it tries to use a tool and fails, enabling the model to make multiple attempts at accomplishing the task and increasing the overall success rate.
Command R models are designed to enhance productivity by seamlessly integrating generative AI capabilities into everyday apps and workflows. Businesses can now streamline their processes and improve overall efficiency, leading to better business outcomes. With Command R+, enterprises can unlock new possibilities and elevate employee and customer experiences.
Cohere instills robust data privacy measures, allowing customers to retain complete control over their data. From customization to model inputs and outputs, businesses can rest assured that their sensitive information remains secure and under their supervision.

Meet Cohere's Command FM

Command is a text generation model for business use cases.

Use cases

Craft your message with an AI assistant so you can write more clear and succinct emails.

Capture key points from an email chain, financial report, or customer call recording.

Provide users with more relevant and personalized search results through semantic search, designed to match the user intent behind a query.

Ask questions and get answers from your company’s entire knowledge base—from your messaging platform to your cloud storage provider to your CRM. Answers come with citations so you can confirm accuracy.

Input a set of data and have your AI assistant provide you with takeaways.

Model version

Rerank 3.5

Rerank 3.5 enhances search accuracy by reranking keyword and vector results. This ensures that only the most relevant content reaches the model—delivering better responses while reducing both latency and costs.

Max tokens: 4,096

Languages: 100+ languages including English, Chinese, Korean, Hindi, Japanese, Spanish, German, French, Arabic, Russian, and Portuguese

Fine-tuning supported: No

Supported use cases: Search-heavy, document-heavy, and RAG scenarios (for example, searching for a hotel)

Read the blog

Command R+

Command R+ is Cohere's most powerful generative language model optimized for long-context tasks, such as RAG and multi-step tool use.


Max tokens: 128K

Languages: English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Arabic, and Chinese

Fine-tuning supported: No

Supported use cases: Text generation, text summarization, chat, knowledge assistants, Q&A, RAG

Read the blog

Command R

Command R is Cohere's generative language model optimized for long-context tasks, such as RAG and tools and large scale production workloads.

Max tokens: 128K

Languages: English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Arabic, and Chinese

Fine-tuning supported: No

Supported use cases: Text generation, text summarization, chat, knowledge assistants, Q&A, RAG

Read the blog

Command

Command is Cohere’s generative large language model (LLM).

Max tokens: 4K

Languages: English

Fine-tuning supported: Yes

Supported use cases: Chat, text generation, text summarization

Read the blog

Command Light

Command Light is a smaller version of Command, Cohere's generative LLM.

Max tokens: 4K

Languages: English

Fine-tuning supported: Yes

Supported use cases: Chat, text generation, text summarization

Read the blog

Embed 3 Multilingual

Embed 3 is Cohere's advanced text and image representation, or embeddings, model. This version supports 100+ languages and delivers exceptional performance for cross-lingual semantic search and retrieval tasks.

Max tokens: 1,024

Languages: Multilingual (100+ supported languages)

Fine-tuning supported: No

Supported use cases: Semantic search, RAG, classification, clustering, multimodal search and retrieval

Read the blog
Learn more about multimodal embeddings

Embed 3 English

Embed 3 is Cohere's advanced text and image representation, or embeddings, model. This version supports English only and delivers exceptional performance for semantic search and retrieval tasks.

Max tokens: 1,024

Languages: English

Fine-tuning supported: No

Supported use cases: Semantic search, retrieval-augmented generation (RAG), classification, clustering, multimodal search and retrieval.

Read the blog
Learn more about multimodal embeddings