Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders.<\/em> Subscribe Now<\/em><\/p>\n\n\n\n I was in more meetings than usual today so I just caught up to the fact that Cohere, <\/strong>the Canadian startup geared co-founded by former Transformer paper author Aidan Gomez <\/strong>toward making generative AI products work easily, powerfully, and securely for enterprises, has released its first reasoning large language model (LLM), Command A Reasoning.<\/strong><\/p>\n\n\n\n It looks to be a strong release. Benchmarks, technical specs, and early tests suggest the model delivers on flexibility, efficiency, and raw reasoning power. <\/p>\n\n\n\n Customer service, market research, scheduling, data analysis are some of the tasks Cohere says it\u2019s built to handle automatically at scale inside secure enterprise environments.<\/strong><\/p>\n\n\n\n It is a text-only model, however, but it should be easy enough to hook up to multimodal models and tools. In fact, tool use is one of its primary selling points. <\/p>\n\n\n\n Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:<\/p>\n\n\n\n Secure your spot to stay ahead<\/strong>: https:\/\/bit.ly\/4mwGngO<\/p>\n\n\n\n While it\u2019s open for researchers to use for non-commercial purposes, enterprises will need to pay Cohere to get access and the company doesn\u2019t publicly list its pricing because it says it makes bespoke customization and private deployment.<\/p>\n\n\n\n Cohere was valued at $6.8 billion<\/strong> when it announced its latest funding round of $500 million a week and a day ago.<\/p>\n\n\n\n Command A Reasoning is tuned for enterprises with sprawling document libraries, long email chains, and workflows that can\u2019t afford hallucinations. <\/p>\n\n\n\n It supports up to 256,000 tokens<\/strong> on multi-GPU setups, a decent size and comparable to OpenAI\u2019s GPT-5.<\/p>\n\n\n\n The research release weighs in at 111-billion parameters, trained with tool-use and multilingual performance in mind. <\/p>\n\n\n\n It supports 23 languages out of the box, including English, French, Spanish, Japanese, Arabic, and Hindi.<\/strong> That multilingual depth is key for global enterprises that need consistent agent quality across markets.<\/p>\n\n\n\n The model slots directly into North<\/em>, Cohere\u2019s new platform for deploying AI agents and automations on-premises. <\/p>\n\n\n\n That means enterprises can spin up custom agents that live entirely within their infrastructure, giving them control over data flows while still tapping into advanced reasoning.<\/p>\n\n\n\n Cohere looks like it\u2019s thought cleverly to identify some of the recurring functions across enterprises \u2014 onboarding, market research and analysis, development \u2014 and trained its model to support its agentic workflows for handling these automatically. <\/p>\n\n\n\n As with many other recent reasoning releases including Nvidia\u2019s new Nemotron-Nano-9B-v2, Command A Reasoning introduces a token budget feature to let users or developers specify how much reasoning to allocate to specific inputs and tasks. Less budget means faster, cheaper replies. More budget means deeper, more accurate reasoning.<\/p>\n\n\n\n The Hugging Face release even exposes this tradeoff directly: reasoning can be toggled on or off through a simple parameter. <\/p>\n\n\n\n Developers can run the model in \u201creasoning mode\u201d for maximum performance or switch it off for lower latency tasks\u2014without changing models.<\/strong><\/p>\n\n\n\n So how does it perform in practice? Cohere\u2019s benchmarks paint a clear picture.<\/p>\n\n\n\n On enterprise reasoning tasks, Command A Reasoning consistently outpaces peers like DeepSeek-R1 0528, gpt-oss-120b, and Mistral Magistral Medium.<\/strong> <\/p>\n\n\n\n It handles multilingual benchmarks with equal strength, important for global businesses.<\/p>\n\n\n\n The token budget system isn\u2019t just a gimmick. In head-to-head comparisons against Cohere\u2019s previous Command A<\/em> model, satisfaction scores climbed steadily as the budget increased. Even with \u201cinstant\u201d minimal reasoning, Command A Reasoning beat its predecessor. At higher budgets, it pulled further ahead.<\/p>\n\n\n\n The story is the same in deep research. On the DeepResearch Bench<\/em>\u2014which measures instruction following, readability, insight, and comprehensiveness\u2014Cohere\u2019s system came out on top against offerings from Gemini, OpenAI, Anthropic, Perplexity, and xAI\u2019s Grok. The model excelled in turning sprawling questions into reports that are not only detailed but readable, a key challenge in enterprise knowledge work.<\/p>\n\n\n\n\n\n\n\n Beyond benchmarks, the model is wired for action. Cohere trained it specifically for conversational tool use \u2014 letting it call APIs, connect to databases, or query external systems during a task.<\/strong><\/p>\n\n\n\n Developers can define tools via JSON schema and feed them into chat templates in Transformers, making it easier to integrate the model into existing enterprise systems.<\/p>\n\n\n\n That design supports Cohere\u2019s larger bet on agentic workflows: AI systems made up of multiple coordinated agents, each handling a piece of a bigger job. Command A Reasoning is the reasoning engine that keeps those workflows coherent and on task.<\/p>\n\n\n\n Cohere is also pitching safety as a central feature. The model is trained to avoid the common enterprise headache of over-refusal \u2014<\/strong> when an AI rejects legitimate requests out of caution \u2014 while still filtering harmful or malicious content. <\/strong><\/p>\n\n\n\n Evaluations focused on five high-risk categories: child safety, self-harm, violence and hate, explicit material, and conspiracy theories.<\/strong><\/p>\n\n\n\n For companies looking to deploy AI in regulated industries or sensitive domains, this balance is meant to make the model more practical in day-to-day operations.<\/p>\n\n\n\n SAP SE is one of the first major partners to integrate the model. Dr. Walter Sun, SVP and Global Head of AI, said the collaboration will enhance SAP\u2019s generative AI capabilities within the SAP Business Technology Platform. For customers, that means agentic applications that can be customized to fit enterprise-specific needs.<\/p>\n\n\n\n Command A Reasoning is available now on the Cohere platform, and for research use on Hugging Face. <\/p>\n\n\n\n The Hugging Face repository provides open weights for research under a CC-BY-NC license, requiring users to share contact information and adhere to Cohere\u2019s Acceptable Use Policy.<\/strong> <\/p>\n\n\n\n Enterprises interested in commercial or private deployments can contact Cohere\u2019s sales team for bespoke pricing.<\/p>\n\n\n\n For enterprises, the pitch is straightforward: one model, multiple modes of deployment, fine-grained control over performance, multilingual capability, tool integration, and benchmark results that suggest it outperforms its peers.<\/p>\n Daily insights on business use cases with VB Daily<\/strong><\/p>\n If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI.<\/p>\n \n\t\t\t\t\tThanks for subscribing. Check out more VB newsletters here.\n\t\t\t\t<\/p>\n An error occured.<\/p>\n<\/p><\/div>\n
\n<\/div>
\n\n\n\n\n
\n<\/div>Tuned for enterprises<\/h3>\n\n\n\n
![\"\"](\"https:\/\/venturebeat.com\/wp-content\/uploads\/2025\/08\/2cf9ae44f8b085a95c125c4bcf4f33deeb1a0ddf-1128x1128-1.webp?w=600\")
<\/figure>\n\n\n\nControlled thinking<\/h2>\n\n\n\n
Excels at enterprise targeted benchmarks<\/h2>\n\n\n\n
Safety: built for high-stakes work<\/h2>\n\n\n\n
Early buy-in from large enterprises<\/h2>\n\n\n\n
Availability and licensing<\/h2>\n\n\n\n
![\"\"\/](\"https:\/\/venturebeat.com\/wp-content\/themes\/vb-news\/brand\/img\/vb-daily-phone.png\")
\n\t\t\t\t<\/div>\n<\/p><\/div>\n<\/div>\t\t\t<\/div>\r\n