Fujitsu and Cohere have revealed plans to create AI models catering specifically to Japanese enterprises. This alliance intends to produce secure, private AI solutions that meet the varied requirements of sectors such as finance, government, and R&D.
Foundation in Command R+
Central to this effort is Cohere’s Command R+ model, recognized for its multilingual accuracy, automated tools, and cost efficiency. Launched in April, this model has demonstrated competitive performance against GPT-4 turbo, particularly in multilingual functions and retrieval-augmented generation (RAG), while remaining budget-friendly on Azure.
Cohere is also known for its Aya 23 series of multimodal AI models. It offers comprehensive support for twenty-three languages, encompassing Arabic, Chinese, French, German, and Japanese, alongside a diverse range of others. This expansive linguistic coverage represents a substantial departure from previous models, which predominantly catered to the English language. The development of Aya 23 leveraged the Aya Collection, a comprehensive dataset containing 513 million prompt-completion pairings.
Boosting Enterprise Search Capabilities
Alongside Command R+, Fujitsu plans to incorporate Cohere’s Embed and Rerank models. These AI models are crafted to enhance enterprise search mechanisms and RAG systems, which will aid businesses in improving their data retrieval and information management processes.
Security and regulatory compliance are key focal points in this partnership. Private cloud deployment options will be offered to accommodate organizations under strict regulatory standards, ensuring that the AI solutions can be securely implemented where sensitive data is involved.
Cohere expressed excitement about the collaboration, noting, “The marriage of our cutting-edge AI technology with Fujitsu’s fine-tuning expertise will equip enterprises with advanced Japanese language models to enhance productivity and efficiency.” The potential impact on global businesses through Fujitsu’s reach was also underscored.
Last Updated on November 7, 2024 3:34 pm CET