Updates to the Command R Series

The latest versions of the Command R model series offer improvements across coding, math, reasoning, and latency.

Today, we’re releasing improved versions of Command R and Command R+, our enterprise-grade AI models optimized for business use-cases. The Command R series excels at retrieval-augmented generation (RAG) with citations, multilingual coverage in over 10 languages to support global business operations, and tool use to automate complex workflows.

The latest versions of our models are more efficient, affordable, and deliver a better experience with new features and enhanced performance. This includes advancements across important areas like coding, math, reasoning, and latency. In particular, Command R, our fastest and most efficient model, has demonstrated material gains across the board and is now on par with the prior version of the much larger Command R+.

This release is a part of our continual effort to improve the quality and efficiency of our models while offering the highest levels of security and data privacy for our customers.

Multilingual RAG & Tools

High accuracy with RAG across languages helps mitigate hallucinations to accelerate global enterprise AI adoption. The Command R series offers in-line citations to help users verify the model outputs. These upgraded versions offer new customization abilities to reduce latency and improve the quality of multilingual RAG. This includes a greater ability to perform planning, tool querying, and answering questions in the user’s language without the need for additional prompting.

Structured Data Analysis

The latest versions of the Command R series offer improvements to coding, math, reasoning, and data analysis. This enables key enterprise use cases across industries like finance, SaaS, and consulting to better drive insights from data in various formats, including tabular data.

Ease of Use

The upgraded Command R series offers new features to enhance ease of use and increase customization options. This includes Structured Outputs to improve the accuracy of JSON generations and better instruction-following in preambles with reduced sensitivity to non-semantic prompt changes.

We are also enabling users to select between two safety mode options (strict or contextual) to balance functionality with appropriate guardrails for enterprise use, depending on the particular use case and sector.

Pricing and Availability

Developers and enterprises can access the latest versions of the Command R series at a more affordable price point on Cohere’s hosted API, Amazon Sagemaker, and OCI's Generative AI service today and on additional cloud platforms soon.

Find additional technical information in our release notes.