Command R is a family of large language models from Cohere built specifically for enterprise use, with a focus on retrieval-augmented generation (RAG), tool use, and multilingual tasks. According to Cohere’s documentation, the models support a 128,000-token context window and are optimized for roughly ten languages including English, French, Spanish, Japanese, and Chinese. The line includes the larger, higher-capability Command R+ alongside Command R.
The models are designed around grounded generation: given a list of supplied document snippets, Command R produces answers with citations indicating which source each claim came from, which is central to building trustworthy business applications on top of a company’s own data. They also have conversational tool-use capabilities, generating structured (JSON) action lists to call external tools and APIs. Cohere updated the family in August 2024 (the command-r-08-2024 identifier) with improved instruction-following and tool-use decisions, and released open weights for the Command R models on Hugging Face for research use.
Command R matters as one of the clearer examples of a model line aimed at enterprise RAG and agentic workflows rather than general chat. For organizations, built-in citation and tool-use behavior reduces the engineering needed to deploy AI that answers from internal documents while showing its sources.