Cohere has launched Embed 4, a new multimodal embedding model designed for enterprise AI applications, including search engines and AI assistants. Embed 4 supports over 100 languages and can process documents with a context length of up to 128,000 tokens, equivalent to around 200 pages. The model accepts both text and image inputs, enhancing its ability to handle complex, unstructured data. Embed 4 is integrated into the Azure AI Foundry model catalog, making it accessible for intelligent document question-and-answer systems, enterprise copilots, and scalable search applications. Additionally, Cohere has become the first model creator to serve its models directly on the Hugging Face Hub, simplifying AI deployment and providing open and enterprise models with OpenAI compatibility. The models are optimized for agentic search, tool use, and multilingual capabilities, targeting enterprise use cases.