Artificial intelligence startup Cohere Inc. today announced the launch of a new large language model, Command R+, designed to support real-world enterprise business workflows and use cases.

Cohere said Command R+ represents the most powerful, scalable LLM developed by the company to date and joins the R-series of models focused on balancing accuracy and efficiency to deliver low cost. It arrives after the Command R, a less capable LLM released last month, and builds on that model’s core strengths to improve performance.

According to the company, Command R+ outperforms similar models in the market based on cost and scalability when applied to enterprise workflow needs on key “business-critical capabilities.” It handles enterprise use cases such as categorization, workflow tool use automation, data analysis and more.

Businesses can readily customize Command R+ with their own proprietary data and it has been optimized for advanced retrieval-augmented generation. RAG improves the accuracy of LLMs by using authoritative knowledge bases outside of training data sources when generating responses. That allows Command R+ to provide highly accurate responses and greatly reduces the potential for hallucinations, Cohere said.

Similar to the Command R, the new model features a 128,000 token context window that allows for the ingestion of extremely large documents and has multilingual coverage of over 10 languages including English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Arabic and Chinese.

One big advance is that Command R+, and its family of large language models, can act as the core reasoning brain for making decisions and doing reasoning for automation of business workflows. When combined with Cohere’s Tools, it can keep customer relationship management tasks, activities and records up to date automatically, assist with customer service, complete orders and perform other duties.

The model also can now support multistep tool use to combine multiple tools over multiple steps to complete complex tasks. The model is even capable of reasoning around issues when a particular tool fails or creates a bug, allowing the model to attempt to correct the path or make multiple attempts to accomplish the task through different means. As a result, the LLM can increase the likelihood of success.

Cohere also announced a new collaboration with Microsoft Corp. to accelerate enterprise AI adoption. Developers and enterprise businesses interested in the Command R+ model can access it on Microsoft Azure starting today. The model will soon become available on Oracle Cloud Infrastructure and additional cloud platforms in the coming weeks.

Meanwhile, developers can get hands-on experience with the newest model in Cohere’s demo environment to test it in a chat environment.

Image: geralt/Pixabay