CohereCommand

Command A+

Command A+ combines image understanding, multilingual text generation and tool use in one model. Its sparse architecture has 218 billion total parameters and activates 25 billion. It expands Command coverage to 48 languages.

Capabilities and access

Thinking can be enabled or disabled. The public API is rate limited; Cohere directs production use to Model Vault. The model is also released as downloadable weights. These routes allow separate decisions about model behavior and where inference runs.Command A+ documentation

When this model fits

Practical guidance. Consider A+ when one task combines a chart, written instructions and calls to business tools. Review the complete workflow, including how a failed tool call is handled. Use a smaller thinking budget for routine routing and compare it with a larger budget on ambiguous cases. A large output allowance is useful for substantial reports, but clear evidence and a requested length still matter.

Specifications & access

Command A+ specifications
DetailValue and conditions
AccessAPI, Downloadable
Model identifierscommand-a-plus-05-2026
Context window128,000 tokens

Cohere API model limit; input and generated output must fit the endpoint's context rules.

Command A+ documentation
Maximum output64,000 tokens

Maximum generated output reported by Cohere; request limits can reduce this.

Command A+ documentation
Inputs and outputText and images in; text outCommand A+ documentation
Weight licenseApache 2.0Command A+ documentation

Related models

  • Coding model with 30 billion total parameters and three billion active parameters for repository and terminal tasks.

    Launch date not confirmedAPI · Downloadable
  • A 111-billion-parameter reasoning model with controlled thinking and a 256,000-token context across 23 languages.

    Launch date not confirmedAPI · Downloadable
  • Vision-language model for documents, charts, tables and visual questions, with up to 20 images per request.

    Launch date not confirmedAPI · Downloadable