Introduction

Today, we’re introducing Command A Reasoning, our most advanced model for enterprise reasoning tasks. From agentic workflows to end-to-end systems, it outperforms leading privately deployable models in its class, including gpt-oss-120b, DeepSeek-R1 0528, and Mistral Magistral Medium. Command A Reasoning is purpose-built for enterprise needs, offering highly secure, efficient, and scalable deployment options. For low footprint deployments, it can run on a single H100 or A100 with a context length of 128k. For latency optimized deployments on two or more GPUs the context length scales to 256k. This configurability ensures organizations can make the most of the hardware available to them. The model's long context length makes it ideal for document-heavy workflows and complex multi-step agentic use cases.

Training

Parameters