> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zerotwo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Cohere

> Cohere models in ZeroTwo — enterprise-grade NLP with industry-leading document understanding and RAG capabilities.

Cohere is a Canadian AI company specializing in enterprise natural language processing, with a particular focus on document understanding, retrieval-augmented generation (RAG), and grounded, factual responses. Cohere models are available in ZeroTwo for users who need reliable, citation-grounded answers from large document collections.

***

## Available Models

| Model       | Type | Strengths                                                                 | Plan Required        |
| ----------- | ---- | ------------------------------------------------------------------------- | -------------------- |
| Command A   | Text | Enterprise flagship — document analysis, long context, grounded responses | Pro+ (Premium)       |
| Command R   | Text | RAG-optimized with citations, grounded factual responses                  | Pro+                 |
| Command R+  | Text | Enhanced Command R with stronger reasoning and instruction-following      | Pro+ (Premium)       |
| Command R7b | Text | Lightweight, fast, efficient — standard model                             | Standard (all plans) |

***

## Strengths

**Retrieval Augmented Generation (RAG):** Cohere's Command R models are purpose-built for RAG workflows — they're optimized to stay grounded in provided documents and cite their sources, rather than generating content from general training knowledge. This makes them excellent for document Q\&A use cases.

**Enterprise document workflows:** Command A is designed for enterprise use cases involving large document collections, knowledge bases, and structured data extraction.

**Grounded factual responses:** Cohere models are specifically trained to stay close to provided context and avoid hallucination. They're a strong choice when factual accuracy is paramount and you're providing reference material.

**Embedding and reranking:** Cohere is also known for its embedding and reranking models (used in backend RAG pipelines). While ZeroTwo surfaces the chat models, Cohere's strength in the retrieval layer translates to better grounding in chat responses.

**Long context:** Command A supports long context windows, allowing it to process and reason across large documents without losing track of earlier content.

***

## Best Use Cases

<CardGroup cols={2}>
  <Card title="Document Q&A" icon="file">
    Upload a contract, report, or knowledge base and ask targeted questions. Cohere excels at staying grounded in the document rather than drifting to general knowledge.
  </Card>

  <Card title="Enterprise knowledge retrieval" icon="search">
    Querying internal knowledge bases, policies, and structured information repositories.
  </Card>

  <Card title="Factual research" icon="book">
    Tasks requiring precise, citation-backed answers with minimal hallucination.
  </Card>

  <Card title="Data-grounded analysis" icon="layers">
    Analyzing structured data, reports, and multi-document summaries with reliable attribution.
  </Card>
</CardGroup>

***

## Choosing a Cohere Model

| If you need...                   | Use...                  |
| -------------------------------- | ----------------------- |
| Best overall document analysis   | Command A               |
| Grounded RAG with citations      | Command R or Command R+ |
| Stronger reasoning on top of RAG | Command R+              |
| Fast, efficient standard model   | Command R7b             |

***

<Tip>
  When using Cohere models for document analysis, upload the relevant documents first and then ask your questions. Cohere's models are specifically trained to prioritize provided context over general knowledge — this is their key advantage.
</Tip>