Cohere has updated its Command family with Command A Plus, a move that keeps the company squarely aimed at enterprise buyers who want performance without giving up control.
The new model appears on Cohere's documentation as command-a-plus-05-2026, and the timing matters. Enterprise AI is no longer just a race to the biggest model or the flashiest demo. It is now a competition over where the model runs, how well it handles retrieval and tools, and whether buyers can keep sensitive data inside systems they trust.
That is the lane Cohere has been trying to own for years. Its public positioning leans heavily on private deployment, RAG, citations, tool use, multilingual performance and infrastructure flexibility, which is a sharper pitch for regulated industries than the one-size-fits-all story offered by the biggest API vendors. On its Command page, Cohere says the family is built for secure, production-ready agentic work and can be deployed privately or inside a hyperscaler VPC, which is exactly the sort of language Fortune 500 security teams want to hear.
Command A Plus sits inside that broader enterprise strategy. Cohere's documentation lists the API model ID as command-a-plus-05-2026, while describing Command A as its most performant model to date, with 111B parameters, a 256,000 token context window and support for tool use, agents and retrieval-augmented generation. That makes this look less like a clean break from the Command A line and more like a continuing refresh of the company's flagship enterprise model.
What stands out is not just scale, but efficiency. Cohere says Command A delivers 150% higher throughput than Command R+ 08-2024 while still running on two A100 or H100 GPUs. That matters because enterprise buyers are increasingly looking past benchmark theater and asking what a model costs to run in production, especially when the use case is high-volume support, document work, research assistants or internal copilots.
The model is also described as strong in 23 languages and optimized for conversational work, long inputs and financial extraction. Those are practical features, not marketing flourishes. A bank, insurer or multinational manufacturer does not care much about abstract model charisma. It cares whether the system can read a policy, pull the right citation and stay accurate when the prompt gets messy.
Why Cohere keeps leaning into control
Cohere's core argument is that enterprise AI needs more than a public API. Its own marketing emphasizes secure deployment, customization and integration with internal data, and that lines up with the company's long-running focus on regulated customers that need data residency and auditability. In that respect, Command A Plus is less about winning the consumer imagination and more about making procurement teams comfortable enough to put the model into real workflows.
That positioning is becoming more valuable as buyers face a widening choice set. Open-weight models from vendors such as Mistral, and a fast-moving wave of enterprise-capable systems from larger players, are making it harder for one company to claim the center of the market. Alibaba's Qwen 3 family has added more pressure in the open-weight arena, and the result is a mid-2026 market where deployment flexibility may matter as much as raw model quality.
For Cohere, that is both the risk and the opportunity. If a customer wants complete control, self-hosting and tighter data governance, the company has a clear message. If a customer wants the broadest ecosystem or the lowest friction at the front door, the larger API platforms still have scale on their side. Command A Plus is Cohere's answer to the buyers sitting in the middle, the ones who want something powerful but do not want to hand over the keys.
The distribution strategy matters because technical buyers increasingly want more than a hosted endpoint. Cohere's wider model lineup is available across its own platform, major cloud services and private deployment options, giving customers room to test, integrate and govern the technology in environments they already use. That does not remove the need for careful evaluation, but it makes the sales pitch more concrete for teams that have to answer to security, legal and finance before anything goes live.
That is why this launch lands as more than a routine model refresh. It reflects where enterprise AI is heading now. The next wave will not be won only by bigger context windows or faster token generation. It will be won by the vendors that can make a strong case for security, deployment control and usefulness inside real companies, and Cohere is making a very deliberate bet that those are the problems worth solving.
Also read: Stability AI pushes deeper into music as AI audio races toward licensing • Samsung defuses a strike threat that could have jolted the AI chip chain • Trump's new crypto order could open the door for fintech startups
