Models

Parse (Details and Application)

Parse is a vision language model that converts unstructured data from enterprise documents to structured outputs that can be used for AI knowledge applications. It provides users with one of the most attractive price-performance options available, making it the ideal choice for high-volume enterprise parsing workloads. Parse can be used alongside Embed and Rerank to build search and retrieval pipelines, or as part of Compass.

Overview

  • Latest model: parse-v5.0
  • Context Length: 8192
  • File types supported: PDF, PPT, JPEG (Base64 encoded)
  • Output format: Markdown
  • Number of Parameters: 2.3B
  • Model Size: ~4.6GB
  • Model Architecture: Proprietary north-micro-vision-instruct
  • Endpoint: Parse
  • API reference: Parse

What it extracts

  • Text and reading order
  • Tables
  • Lists
  • Forms and key-value pairs
  • Images and captions
  • Locations of page boundaries and visual elements

Output

Returns Markdown containing:

  • Document text and content
  • Lists
  • Tables represented as HTML
  • Bounding box coordinates
  • Image descriptions
  • Content that can be rendered directly as formatted Markdown/HTML

Supported languages

Parse is stable for nine input langauges. The model supports zero-shot parsing of additional languages, though accuracy may be lower than for the recommended languages.

ISO CodeLanguage Name
arArabic
enEnglish
frFrench
deGerman
jaJapanese
koKorean
itItalian
ptPortuguese
esSpanish

Best for

Parse is particularly well suited for the following use case areas where throughput, scale and cost effectiveness are key performance considerations:

  • Preparing documents for search and retrieval
  • RAG ingestion pipelines
  • Intelligent document processing
  • Providing document context to AI agents

Known limitations

The model does not return confidence scores for extracted content. The model does not identify document elements such as headers, footers, or font hierarchy. Parsed content is returned as Markdown; structured JSON output is not supported.

Getting started