Generative artificial intelligence model startup Cohere Inc. today announced it has raised $500 million in a late-stage funding round, bringing the company’s valuation to $5.5 billion.

The Series D funding round was led by PSP Investments, a Canadian pension investment manager, and new backers that include Cisco Systems Inc., Japan’s Fujitsu Ltd., chipmaker Advanced Micro Devices Inc.’s AMD Ventures and Canada’s export credit agency EDC.

Reuters previously reported on the deal, which doubles the company’s valuation from $2.2 billion last year when the company raised $270 million from investors including Nvidia Inc., Oracle Corp. and Magnetar Capital.  This funding round brings the total raised by the company to $970 million.

Cohere builds enterprise-focused AI large language models and stands as a rival to OpenAI and Google LLC, which also dominate the industry. There are very few startups in the industry in part because of how expensive and difficult it is to develop and deploy LLM foundation models. As a result, there are not many experts and resources available to do the research and development needed to stay ahead.

The company’s approach has been to build practical foundation models that can be implemented by enterprise customers to support employees and automate business systems. Unlike other high-profile companies in the industry, which are pursuing “human-like” intelligence in the form of AI, Cohere aims to build AI for enterprise efficiency.

Cohere’s most recent AI model, Command R+, continues this vision by aiming to support real-world enterprise business workflows and use cases. The company’s Command R family of models is designed to do generative AI business tasks, such as summarizing and analyzing text, but also have the capability of applying reasoning to automate software tools. This means that Command R+ can be commanded to complete multi-step tasks and it will combine and use available software to complete those tasks.

The company’s models are multilingual and cover 10 languages, including English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Arabic and Chinese.

With the power of generative AI foundation models, customers can interface with software and data in ways that were never possible before. LLMs make it possible to “talk” to data, by holding human-like conversations, at the same time the model can understand intent and context from both data and users. This makes it much easier for a nontechnical user to approach difficult tasks that might otherwise require expert knowledge to understand.

Cohere’s customers include a broad range of industries, which use its models to summarize, analyze and explain data from within finance, technology and retail. In some cases, its models are used in retail applications to recommend products and in banking applications to answer questions based on financial documents.

The company recently partnered with Fujitsu to develop LLMs specifically tailored for enterprise use cases with Japanese language capabilities, tentatively named “Takane,” based on Command R+. The joint partnership will lead to fine-tuned models for specialized model development and private cloud deployments for highly regulated industries such as finance, healthcare and government agencies.

Image: Pixabay