Cohere today unveiled Embed 3, its most advanced multimodal AI model, which seamlessly integrates text and image embeddings within a unified latent space, setting new benchmarks for accuracy and performance in enterprise search and multilingual retrieval tasks.
The model is capable of generating embeddings from both text and images enabling businesses to unlock valuable insights from their vast data, including complex reports, product catalogs, and design files, boosting workforce productivity.
Embed 3 is now available on Cohere’s platform, Amazon SageMaker, and for private deployment in any VPC or on-premise environment.
Beyond Multimodal Capabilities
Embed 3 converts data into numerical representations within a unified vector space, allowing for accurate similarity comparisons across text and image data. This ensures balanced and highly relevant search results without bias toward one modality, setting it apart from other models.
Embed 3 excels in various real-world use cases. For instance, businesses can now retrieve graphs, charts, and eCommerce product images more efficiently, enhancing data-driven decision-making and customer experience. It also simplifies the creative process for designers by allowing quick searches for UI mockups and visual templates based on text descriptions.
Embed 3 now supports over 100 languages, making it an essential tool for global enterprises.
Use cases of Embed 3’s multimodal AI search in Enterprise are huge.
For example, users can easily find relevant graphs and charts by describing specific insights, making data-driven decision-making more efficient across teams. In e-commerce, Embed 3 transforms the search experience by allowing customers to search both product images and text, improving the shopping experience and boosting conversion rates. Designers also benefit, as they can quickly locate UI mockups and visual templates with text descriptions, streamlining the creative process and reducing the time spent searching through large asset libraries.
Founded in 2019, Cohere specialises in developing large language models (LLMs) designed for business applications. Unlike some of its counterparts, such as OpenAI and Google, Cohere focuses on enterprise solely rather than pursuing artificial general intelligence (AGI).
Earlier this year, in June, Cohere reached a valuation of $5.5 billion, solidifying its position as one of the world’s most valuable AI companies and one of Canada’s largest startups. They raised $500 million in their series D funding round. Clients like Notion Labs and Oracle use the company’s models to streamline website copywriting, user communication, and generative AI integration.
