Cohere announces Parse 5, a model that turns documents into Markdown
The 2.3-billion-parameter vision language model converts PDF, presentation and image pages into structured text. The company's price is $1.50 per 1,000 pages.
Cohere announced Parse 5, a model for processing enterprise documents. The company presents it as an option that keeps per-page cost low while maintaining parsing accuracy, aimed at high-volume enterprise work.
It is a 2.3-billion-parameter vision language model with an 8,192-token context window and a footprint of roughly 4.6 GB, built on Cohere Labs' North-Micro-Vision-Instruct architecture.
It works like this: a PDF, presentation or JPEG page is supplied as a base64-encoded data URI, and the model returns Markdown. The output orders text by reading sequence; tables come back as HTML, and lists and form key-value pairs are marked separately. Images get descriptions along with bounding box coordinates.
The stated price is $1.50 per 1,000 pages. Cohere argues this is among the lowest per-page rates in the sector; that comparison is the company's own assessment and has not been independently verified.
Parse 5 is available through the Cohere Parse API, Microsoft Foundry, AWS SageMaker and the single-tenant Model Vault. The company also says it has opened a Hugging Face Space where the model can be tried for free, and that Parse is available to North and Compass users.
For details see Cohere's documentation.
