Source URL: https://simonwillison.net/2025/Mar/13/command-a/#atom-everything
Source: Simon Willison’s Weblog
Title: Introducing Command A: Max performance, minimal compute
Feedly Summary: Introducing Command A: Max performance, minimal compute
New LLM release from Cohere. It’s interesting to see which aspects of the model they’re highlighting, as an indicator of what their commercial customers value the most (highlight mine):
Command A delivers maximum performance with minimal hardware costs when compared to leading proprietary and open-weights models, such as GPT-4o and DeepSeek-V3. For private deployments, Command A excels on business-critical agentic and multilingual tasks, while being deployable on just two GPUs, compared to other models that typically require as many as 32.
It’s open weights but very much not open source – the license is Creative Commons Attribution Non-Commercial and also requires adhering to their Acceptable Use Policy.
Cohere offer it for commercial use via their API. I released llm-command-r 0.3 adding support for this new model, plus their smaller and faster Command R7B (released in December) and support for structured outputs via LLM schemas.
(I found a weird bug with their schema support where schemas that end in an integer output a seemingly limitless integer – in my experiments it affected Command R and the new Command A but not Command R7B.)
Via @Prince_Canuma
Tags: llm, structured-extraction, cohere, generative-ai, ai, llms
AI Summary and Description: Yes
Summary: The text discusses the launch of Command A, an LLM (large language model) from Cohere that promises high performance with minimal hardware requirements, making it a noteworthy development in the generative AI space. Its focus on cost efficiency and capability for multilingual tasks positions it as a competitive option for commercial use.
Detailed Description:
The introduction of Command A by Cohere signifies a notable advancement in the realm of large language models, particularly focusing on efficiency and performance with limited hardware resources. Security and compliance professionals in AI and cloud computing should take note of its implications for deployment and operational costs.
– **Performance and Hardware Costs**:
– Command A is designed to deliver maximum performance using only two GPUs, which is a significant reduction compared to leading models that often require 32 GPUs.
– This efficiency could make it easier for organizations to adopt advanced AI capabilities without substantial upfront hardware investments.
– **Deployment and Use Cases**:
– Targeted towards multilingual and agentic tasks, Command A is poised to be particularly beneficial for businesses reliant on diverse language support.
– The model’s ability to function effectively in private deployments underscores its relevance for organizations concerned about data privacy and security.
– **Licensing**:
– While the model utilizes open weights, it is not open source. The licensing is under Creative Commons Attribution Non-Commercial, indicating restrictions on its commercial deployment unless through their API, which highlights the need for professionals to be aware of licensing implications for compliance.
– There are also constraints related to adherence to their Acceptable Use Policy, which can influence deployment strategies and applications in sensitive environments.
– **Accompanying Tools**:
– The release also mentions the support for structured outputs via LLM schemas, which are critical for organizations looking to integrate AI outputs into structured data systems.
– An issue was noted with the schema support that may raise concerns about reliability and output management in practical applications.
– **Commercial Strategy**:
– Cohere’s approach to providing API access emphasizes the shift towards cloud-based AI solutions, aligning with current trends in software and infrastructure security where API vulnerabilities can pose risks.
In conclusion, the launch of Command A reflects robust advancements in generative AI capabilities while highlighting considerations for security and compliance professionals regarding deployment efficiency, licensing, and potential risks associated with schema issues. This development will likely influence how businesses approach AI adoption, particularly in cloud environments.