Cohere Parse

#19 in Multimodale Modelle

Cohere · 2× · tolest 29. Aug. 2026

32

Momentum

Cohere Parse is a Vision-Language Model from Cohere that transforms complex enterprise documents (PDFs, presentations, images) into structured, machine-readable Markdown outputs. Beyond plain text, it recognizes tables, forms, diagrams, and embedded images including bounding boxes, and is designed for cost-effective processing of large document volumes in enterprise environments. The model (parse-v5.0) has 2.3 billion parameters, a context window of 8,192 Token, and is available via the Cohere API, Microsoft Foundry, AWS SageMaker, and as a dedicated Model Vault deployment.

Belege (2)