Models·Cohere

Cohere Command A+ Model Review

cohere/command-a-plus

Cohere’s text-and-image model for reasoning and tool-assisted tasks.

$0.30Input, per million tokens
$1.50Output, per million tokens
192KContext window
64KLongest answer

192K tokens of context — about 144,000 words, on a scale that ends at 2M.

What it can do

  • Reasoning
  • Vision
  • Tool use
  • Takes text
  • Takes image
  • Gives text

Released 22 Sept 2026 · checked 3 Oct 2026.

About it

Command A+ is Cohere’s model for tasks that use reasoning and tools. It accepts text and images, and returns text. Its 192,000-token context window lets it take in lengthy material, while its maximum answer is 64,000 tokens. Input costs $0.3 per million tokens; output costs $1.5 per million tokens.

Review

In a line

Command A+ is Cohere’s model for reasoning and tasks that use tools. It accepts text and images and produces text.

What it's good at

The model is suited to reasoning tasks and can call tools using defined rules for how those tools should be used. It can also take images as input. See other models with tool calling and Cohere models.

Price in plain terms

Input costs $0.3 per million tokens, and output costs $1.5 per million tokens. A token is a small piece of text the model processes or produces; images are also accepted, but no image-specific price is listed.

At these rates, reading a 500-word page costs about 0.019 cents. A chat turn with about 1,000 input tokens and 500 output tokens costs about 0.1 cents. One thousand such turns cost about $1.05.

Context and limits

Command A+ can take in up to 192,000 tokens at once. That is the context window: the amount of material it can consider in a request. Its maximum answer is 64,000 tokens.

It accepts text and images but returns text.

Ways to use it

You can provide text or images and receive a text response. The model supports native tool calling, with strict rules for the information passed to tools.

Verdict

Consider Command A+ if you need a model that can reason, handle images, and use tools, with room for a large amount of input and a long answer. Its per-token prices are low in the examples provided, though output costs more than input. If you need image output or a specific app connection, the listed capabilities do not establish that this model provides either.

Verdict

Command A+ may suit people choosing a model for reasoning or tool-assisted work, especially when their requests include images or a large amount of text. Its large context and answer limits allow long requests and responses. Budget for both input and output: output tokens cost more, and frequent or lengthy answers add to the bill.

  • Accepts both text and image inputs.
  • Supports reasoning and native tool calling.
  • Has a 192,000-token context window.
  • Can produce answers up to 64,000 tokens.
  • Example costs are about $1.05 for 1,000 chat turns.
  • Returns text only.
  • Output tokens cost five times as much as input tokens.
  • No specific app connections or API providers are listed.
  • Pick it ifReasoning and tool-assisted tasks; Requests that include text and images; Long inputs or long text answers
  • Skip it ifYou need the model to produce images.; You need a confirmed connection to a particular app.

Pricing

Input is $0.3 per million tokens and output is $1.5 per million tokens. The example costs are about 0.019 cents to read a page, about 0.1 cents for one chat turn, and about $1.05 for 1,000 such turns.

Questions

How much does Command A+ cost?

Input costs $0.3 per million tokens; output costs $1.5 per million tokens.

What does a typical chat turn cost?

A turn with about 1,000 input tokens and 500 output tokens costs about 0.1 cents. A thousand such turns cost about $1.05.

Can Command A+ understand images?

Yes. It accepts images as input and returns text.

How much text can it handle at once?

Its context window is 192,000 tokens, which is the amount of material it can consider in one request.

How long can its answer be?

It can produce up to 64,000 tokens in one answer.

Can it use tools?

Yes. It supports native tool calling with strict rules for tool use.

On these boards

More from Cohere