Cohere · Available now

Command A+ (Cohere, 2026): price per reply, context window and Apache 2.0 licence

Command A+ is Cohere’s open-weights model (218B parameters, 25B active, Apache 2.0, 128K-token input context), released on May 20, 2026, and OpenRouter lists it at $0.30 per million input and $1.50 per million output tokens.

Available nowLast updated:

Use it now

Chat

≈ 1 credit per reply · See all models

New accounts get 5 free credits once, valid for 7 days. They work on the lower-priced models here; the priciest ones need credits you bought.

Start chatting — 5 free credits
Maker
Cohere
Type
Text
Released
2026-05-20
OpenRouter id
cohere/command-a-plus

Per million tokens

Price

Command A+ · checked 2026-11-17
-InputOutputSource
Via OpenRouter (cohere/command-a-plus)$0.30$1.5openrouter.ai

Checked

Facts, with sources

  • Cohere’s launch post is dated May 20, 2026 and releases Command A+ as an open-source model. Cohere’s changelog says it is "now available for all Cohere users through our standard API endpoints".Source: cohere.com · checked 2026-11-17
  • Cohere calls Command A+ "our fastest and most powerful language model yet", an enterprise model for complex reasoning and multimodal, multilingual agentic tasks. Its changelog calls it the last model in the Command A family.Source: docs.cohere.com · checked 2026-11-17
  • Sparse mixture-of-experts model with 218B total and 25B active parameters (the Hugging Face card lists 128 experts, 8 active per token, plus one shared expert), a 128K-token input context and up to 64K output tokens. Input is text and images; output is text. It supports 48 languages and a reasoning ("thinking") mode.Source: huggingface.co · checked 2026-11-17
  • Open weights on Hugging Face in BF16, FP8 and W4A4 versions. Cohere’s card names W4A4 as the recommended one and lists the smallest setup as 2 × H100 or 1 × B200 GPUs for W4A4 (BF16 needs 8 × H100 or 4 × B200). It can run with vLLM, Transformers and other local runners. We have not downloaded or run the weights.Source: huggingface.co · checked 2026-11-17
  • The Hugging Face card and Cohere’s launch post give the licence as Apache 2.0, a standard open-source licence that allows commercial use, modification and redistribution with notice and attribution terms. This is a summary, not legal advice.Source: huggingface.co · checked 2026-11-17
  • Cohere’s pricing page lists Command A+ as "Free" with $0 for "API key" and $0 for "Model download", and gives no per-token price for it. The same page says calls made with a Trial API key are free but rate limited and not allowed for production or commercial use, and that calls made with a Production API key are billed pay-as-you-go. We could not find a Cohere per-token price for Command A+ on Cohere’s own pages.Source: cohere.com · checked 2026-11-17
  • Cohere’s model list shows Command A+ as live, with API model name command-a-plus-05-2026 on the Chat endpoint, 128K context, 64K maximum output, and text and image input. It also gives the Microsoft Azure id coherelabs-command-a-plus-05-2026-w4a4 and the OpenRouter id cohere/command-a-plus. Dedicated inference is offered through Cohere’s Model Vault, and enterprise customers can request private deployment.Source: docs.cohere.com · checked 2026-11-17
  • Listed on OpenRouter as cohere/command-a-plus, a text-and-image-input, text-output model with a 192,000-token total context (128,000 prompt plus 64,000 completion at its only listed host, Cohere, served in FP8) and native tool calling, at $0.30 per million input tokens, $1.50 per million output tokens and $0.15 cached input (checked November 17, 2026).Source: openrouter.ai · checked 2026-11-17 · second-hand
  • Cohere’s launch post says Command A+ combines what earlier Command A variants (Reasoning, Vision, Translate) did into one model, calls it Cohere’s first multimodal reasoning model, raises language support from 23 to 48 languages, and uses a newer tokenizer that produces fewer tokens for the same output. Cohere also publishes its own speed and benchmark comparisons against Command A Reasoning and Command A Vision (the maker’s own figures, not tested by us; we do not repeat them).Source: cohere.com · checked 2026-11-17

Nearest available

Models you can use now instead

$0.30 / $1.2 (per million tokens)

  • DeepSeek released V4.1-Flash on September 10, 2026.Source: api-docs.deepseek.com · checked 2026-11-11
  • Only the Flash version of V4.1 exists: DeepSeek has published no V4.1 Pro, and the Pro model on its price page is still deepseek-v4-pro (V4-Pro-0813).Source: api-docs.deepseek.com · checked 2026-11-11
  • DeepSeek’s pricing page says the API model deepseek-flash is now served by the V4.1-Flash-1017 snapshot.Source: api-docs.deepseek.com · checked 2026-11-11

Qwen4-Max

Alibaba

$1.5 / $4.5 (per million tokens)

  • Alibaba Cloud’s Model Studio page for qwen4-max is labelled "Latest Model", was last updated on November 16, 2026, and calls Qwen4-Max "the next-generation flagship multimodal model in the Qwen family". We found no Alibaba announcement or release date, so none is stated here.Source: alibabacloud.com · checked 2026-11-16
  • Context window 1,000,000 tokens; text, image and video input, text output. Alibaba’s help page lists a maximum output of 131,072 tokens in non-thinking mode, while QwenCloud’s model page shows a larger limit for reasoning; the two pages disagree, so we give no single figure.Source: qwencloud.com · checked 2026-11-16
  • Alibaba’s QwenCloud price page lists qwen4-max at $1.50 per million input tokens and $4.50 per million output tokens (International / Singapore region), the same in thinking and non-thinking modes, with implicit cache reads at $0.075.Source: docs.qwencloud.com · checked 2026-11-16

FAQ

Questions

What is Command A+?

Command A+ is a model from Cohere, released on May 20, 2026 as an open-weights model. It is a sparse mixture-of-experts model with 218B total and 25B active parameters, takes text and image input, produces text, supports 48 languages and has a reasoning mode. Cohere calls it its fastest and most powerful model yet and the last model in the Command A family.

How much does Command A+ cost per reply?

Cohere’s pricing page shows no per-token price for Command A+; it lists the model as $0 for an API key and for the weight download, says Trial API key calls are rate limited and not allowed for production or commercial use, and says Production key calls are billed pay-as-you-go. OpenRouter lists cohere/command-a-plus at $0.30 per million input tokens, $1.50 per million output tokens and $0.15 cached input (checked November 17, 2026). On this site a typical reply to this model costs about 1 credit (a credit is about $0.01 in the $5 pack), billed by its real token cost, at least 1 credit.

How do I use Command A+?

Through Cohere’s API with the model name command-a-plus-05-2026 (the Chat endpoint), through Cohere’s Model Vault for dedicated inference, as private deployment for enterprise customers, on Microsoft Azure, on OpenRouter as cohere/command-a-plus, or by downloading the weights from Hugging Face. You can also chat with it on this page on credits, without a Cohere account.

What is the Command A+ context window?

Cohere lists a 128K-token input context and up to 64K output tokens. OpenRouter shows a 192,000-token total context for it, which is those two limits added together (128,000 prompt plus 64,000 completion).

Is Command A+ open source?

Its weights are on Hugging Face under the Apache 2.0 licence, which allows commercial use, modification and redistribution with notice and attribution terms. Cohere’s card names the W4A4 version as the recommended one, which needs about 2 × H100 or 1 × B200 GPUs. Read the licence text before commercial use.

How is Command A+ different from earlier Command A models?

Cohere says Command A+ puts the abilities of the earlier Command A Reasoning, Vision and Translate variants into one model, is its first multimodal reasoning model, supports 48 languages instead of 23, and uses a newer tokenizer that needs fewer tokens for the same output. Cohere also publishes its own speed and benchmark comparisons; those are the maker’s own figures and we have not tested them.

Is this the official Cohere website?

No. This is an unofficial page, not affiliated with or endorsed by Cohere. Facts link to Cohere’s own pages where possible.