Command R+
Command R+ is a text generation model from Cohere designed for RAG workflows with a 128,000 token context window.
Cohere
Amazon BedrockEnterprise RAG and long-context text generation
Command R+ is a large language model developed by Cohere and released in April 2024. It is available through Amazon Bedrock under the identifier cohere.command-r-plus-v1:0 and is designed primarily for enterprise use cases involving retrieval-augmented generation (RAG) and complex instruction following. The model supports a 128,000 token context window and produces responses up to 4,000 tokens in length.
Command R+ is optimized for tasks that require grounding responses in external documents, making it well-suited for enterprise search, document summarization, and question-answering pipelines. Cohere built the model with multilingual capabilities and tool use in mind, supporting workflows where the model needs to reason over retrieved content or call external APIs. Note that this version is currently marked as deprecated on the Amazon Bedrock platform.
What Command R+ supports
Retrieval-Augmented Generation
Designed specifically for RAG pipelines, allowing the model to ground responses in retrieved documents and reduce hallucinations in knowledge-intensive tasks.
Long Context Window
Supports up to 128,000 tokens of context, enabling processing of lengthy documents, multi-turn conversations, or large knowledge bases in a single request.
Tool Use and Agents
Supports tool use natively, allowing the model to call external APIs or functions as part of multi-step agentic workflows.
Multilingual Text Generation
Generates and understands text across multiple languages, with Cohere reporting support for over 10 languages including French, German, Spanish, and Japanese.
Instruction Following
Trained to follow complex, multi-part instructions accurately, making it suitable for structured enterprise tasks such as summarization, classification, and drafting.
Ready to build with Command R+?
Get Started FreeBenchmark scores
Scores represent accuracy — the percentage of questions answered correctly on each test.
| Benchmark | What it tests | Score |
|---|---|---|
| MMLU-Pro | Expert knowledge across 14 academic disciplines | 43.2% |
| GPQA Diamond | PhD-level science questions (biology, physics, chemistry) | 32.3% |
| MATH-500 | Undergraduate and competition-level math problems | 27.9% |
| AIME 2024 | American math olympiad problems | 0.7% |
| LiveCodeBench | Real-world coding tasks from recent competitions | 12.2% |
| HLE | Questions that challenge frontier models across many domains | 4.5% |
| SciCode | Scientific research coding and numerical methods | 11.8% |
Common questions about Command R+
What is the context window size for Command R+?
Command R+ supports a context window of 128,000 tokens, allowing large volumes of text or document content to be processed in a single request.
What is the maximum response length?
The model can generate responses up to 4,000 tokens in length per request.
Is Command R+ still available on Amazon Bedrock?
The specific version available on MindStudio (cohere.command-r-plus-v1:0) is currently marked as deprecated on Amazon Bedrock. Users should check for updated versions if they require ongoing support.
Does Command R+ support image or video inputs?
No. Based on the available metadata, Command R+ is a text-only model and does not support image or video inputs.
What is Command R+ best suited for?
Command R+ is tagged for RAG (retrieval-augmented generation) and high-performing text tasks. It is particularly well-suited for enterprise search, document question-answering, and agentic workflows that involve tool use.
What people think about Command R+
Community discussion around Cohere's Command R series reflects that the model was widely used as a daily driver by many users roughly two years ago, particularly valued for its RAG capabilities and enterprise focus. Cohere as a company continues to attract attention, having been valued at $6.8 billion in a recent fundraising round.
Some community members have noted uncertainty about the ongoing trajectory of the Command R and Command A model lines, with questions about how they fit into Cohere's current product strategy. Discussions are generally informational rather than critical, with no widespread technical complaints surfacing in the sampled threads.
What ever happened to Cohere’s Command-R and Command-A series of models? R was a lot of folks’ daily driver model like 2 years ago.
GPT -OSS is heavily trained on benchmark. scored rank 34 on simplebench worse than grok 2
AI startup Cohere valued at $6.8 billion in latest fundraising, hires Meta exec
Parameters & options
Explore similar models
CohereCohere Rerank 4 Fast
Cohere Rerank 4 Fast is a reranking model from Cohere designed to score and reorder search results with a 4096 token context window.
CohereCohere Rerank 4 Pro
Cohere Rerank 4 Pro is a reranking model from Cohere designed to improve search result relevance with a 4096-token context window.
CohereCohere Embed 4
Cohere Embed 4 is an embedding model from Cohere with a 128,000-token context window, released in April 2025.
OpenAIGPT 5.6 Luna
GPT 5.6 Luna is a text and image-input model from OpenAI with a 1,050,000-token context window.
DeepSeekDeepSeek V4 Flash 0731
DeepSeek V4 Flash 0731 is a text generation model from DeepSeek with a 1,048,576-token context window and configurable reasoning effort.
GoogleGemini 3.7 Flash
Gemini 3.7 Flash is a text generation model from Google with a 1,048,576-token context window and configurable thinking levels.
Start building with Command R+
No API keys required. Create AI-powered workflows with Command R+ in minutes — free.