Sold by: Cohere
Deployed on AWS
Free Trial
Command A is a highly efficient generative model that excels at agentic and multilingual use cases.
Overview
Command A is Cohere's flagship generative model, optimized for companies that require fast, secure, and highly-performant AI solutions. 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, and can be deployed on just 2 GPUs, compared to competitive models that typically require as many as 32 GPUs. In head-to-head human evaluation across business, STEM, and coding tasks, Command A matches or outperforms its larger and slower competitors, while offering superior throughput and increased efficiency.
Highlights
- Command A is very effective at adapting in real time and solving multiple step problems based on context and the objectives it is given. This unlocks various points of automation and assistance for businesses across the globe.
- Command A is highly balanced. It performs incredibly well at critical uses cases without sacrificing performance in essential areas, making it a very good general model. This balanced starting point also makes Command A particularly apt to being customized and fine tuned to specific use cases as needed.
- Typically such a powerful model would be incredibly expensive to serve, but Command A is incredibly efficient, operating on 2 Ax100s or Hx100s.
Details
Sold by
Delivery method
Latest version
Deployed on AWS
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Free trial
Try this product free for 7 days according to the free trial terms set by the vendor.
Pricing is based on actual usage, with charges varying according to how much you consume. Subscriptions have no end date and may be canceled any time.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
Dimension | Description | Cost/host/hour |
|---|---|---|
ml.g4dn.12xlarge Inference (Batch) Recommended | Model inference on the ml.g4dn.12xlarge instance type, batch mode | $34.25 |
ml.p4de.24xlarge Inference (Real-Time) Recommended | Model inference on the ml.p4de.24xlarge instance type, real-time mode | $34.25 |
Dimensions summary
You pay by the hour for the GPU instance that runs Command A. Two options let you match hardware to your workload. The ml.g4dn.12xlarge instance runs inference in batch mode, where you submit groups of requests to process together. The ml.p4de.24xlarge instance runs inference in real-time mode, where the model responds to requests as they arrive. Each option bills per host hour, so cost scales with how long the instance stays running. You choose the instance and mode that fit your throughput and latency needs.
Top-of-mind questions for buyers
Command A runs on two GPUs. The ml.p4de.24xlarge uses A100-class GPUs for real-time inference. The ml.g4dn.12xlarge provides GPUs for batch inference. The model has 111B parameters and a 256,000-token context window, so it requires two GPUs to run.
You pay per host hour while the instance runs. A stopped instance does not accrue software charges. You control cost by running instances only when needed. Underlying AWS storage or resource fees may still apply while an instance is stopped, separate from the software metering.
The ml.g4dn.12xlarge runs batch inference, where you submit grouped requests to process together. The ml.p4de.24xlarge runs real-time inference, responding to requests as they arrive. Choose batch for scheduled, high-volume jobs. Choose real-time when users or applications need immediate responses. Both bill per host hour.
docs.cohere.com
Helpful?
Vendor refund policy
There are no refunds.
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
Amazon SageMaker model
An Amazon SageMaker model package is a pre-trained machine learning model ready to use without additional training. Use the model package to create a model on Amazon SageMaker for real-time inference or batch processing. Amazon SageMaker is a fully managed platform for building, training, and deploying machine learning models at scale.
Deploy the model on Amazon SageMaker AI using the following options:
Version release notes
API Version Control: Users can now specify the API version (v1 or v2) in the endpoint URL, providing greater flexibility and control over API interactions.
We've updated our SageMaker integration with a major version release for Cohere models, including notebook updates. The "/invocation" endpoint now defaults to API V2, ensuring a seamless transition to the latest version. Please see the notebook on how to use this model with the API update: All existing (and future) Command models: https://github.com/cohere-ai/cohere-aws/blob/main/notebooks/sagemaker/Command%20Models.ipynb
Bug Fixes: -Billing Token Issue: Resolved an issue where billing tokens were consistently returning as 0 for embed requests. -Image Processing Error: Addressed a problem where the inference server failed to process valid base64 image URIs, resulting in "failed to parse image" errors. This issue was specific to the inference server and did not affect other routes.
Additional details
Inputs
- Summary
The model accepts JSON requests with parameters that can be used to control the generated text. See examples and fields descriptions below.
- Input MIME type
- application/json
{ ""messages"" = [ {""role"": ""SYSTEM"", ""content"": ""You are a helpful assistant""}, {""role"": ""USER"", ""content"": ""What is an interesting new role in AI if I don't have an ML background?""}, {""role"": ""ASSISTANT"", ""content"": ""You could explore being a prompt engineer!""}, {""role"": ""USER"", ""content"": ""What are some skills I should have?""}, ], }
N/A
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
message | A list of chat messages in chronological order, representing a conversation between the user and the model. Default value [ ] Messages can be from User, Assistant, Tool and System roles. | - | Yes |
documents | A list of relevant documents that the model can cite to generate a more accurate reply. Each document is either a string or document object with content and metadata. Input data type is text. | - | No |
tools | A list of available tools (functions) that the model may suggest invoking before producing a text response. The input data type is text. Default value is [] When tools is passed (without tool_results), the text content in the response will be empty and the tool_calls field in the response will be populated with a list of tool calls that need to be made. If no calls need to be made, the tool_calls array will be empty. | - | No |
citation_options | Options for controlling citation generation. Input data type is text. | - | No |
logprobs | Defaults to false. When set to true, the log probabilities of the generated tokens will be included in the response. Input data type is categorical. If categorical is chosen then TRUE, FALSE. Default value is FALSE. | - | No |
stop_sequences | A list of up to 5 strings that the model will use to stop generation. If the model generates a string that matches any of the strings in the list, it will stop generating tokens and return the generated text up to that point not including the stop sequence. Input data type is text. | - | No |
strict_tools | When set to true, tool calls in the Assistant message will be forced to follow the tool definition strictly. Note: The first few requests with a new set of tools will take longer to process. | Input data type is categorical. If categorical is chosen then TRUE, FALSE. | No |
tool_choice | Allowed values: REQUIRED, NONE Used to control whether or not the model will be forced to use a tool when answering. When REQUIRED is specified, the model will be forced to use at least one of the user-defined tools, and the tools parameter must be passed in the request. When NONE is specified, the model will be forced not to use one of the specified tools, and give a direct response. If tool_choice isn’t specified, then the model is free to choose whether to use the specified tools or not. | If Categorical is chosen: REQUIRED, NONE | No |
stream | Defaults to false. When true, the response will be a SSE stream of events. Streaming is beneficial for user interfaces that render the contents of the response piece by piece, as it gets generated. | Input data type is Categorical. If Categorical is chosen then TRUE, FALSE. | No |
safety_mode | Allowed values: CONTEXTUAL STRICT OFF Used to select the safety instruction inserted into the prompt. Defaults to CONTEXTUAL. When OFF is specified, the safety instruction will be omitted. Safety modes are not yet configurable in combination with tools, tool_results and documents parameters. | - | No |
Resources
Vendor resources
Support
Vendor support
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Similar products
Command R is a generative language model optimized for long-context tasks and large scale production workloads.
Command R+ is a generative language model optimized for long-context tasks and large scale production workloads.
Command R 08-2024 is a generative language model optimized for large scale production workloads.
Command R+ 08-2024 is a highly performant generative language model optimized for large scale production workloads.
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.