Cohere Parse 5
Cost-effective document parsing at enterprise scale
Context window
8K
Input / 1M tokens
Not announced
Output / 1M tokens
Not announced
Provider
Cohere
Data verified 2026-08-29
Parse 5 is a 2.3-billion-parameter vision language model that converts PDFs, slides, and images into structured Markdown. It extracts tables, forms, images, and preserves document structure without a separate OCR step, optimized for high-volume document processing and cost-effectiveness.
Capability index
Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.
How to access it
Available through Cohere Parse API ($1.50 per 1,000 pages), Microsoft Foundry, AWS SageMaker, and Model Vault (single-tenant deployment).
Strengths
- ✓Exceptional price-to-performance ratio at $1.50 per 1,000 pages
- ✓End-to-end document processing without separate OCR pipeline
- ✓Supports nine major world languages with stable accuracy
- ✓High throughput: 4.5 pages per second per GPU
Best for developers who...
When to choose it (and when not to)
Reach for Cohere Parse 5 when...
- →Processing high volumes of documents where cost efficiency is critical
- →When you need structured data extraction without complex pipelines
Look elsewhere if...
- ✕When peak accuracy on charts or visual grounding is required
- ✕For low-volume processing where general-purpose LLMs may be more appropriate
How to use it
- ›Provide documents as base64-encoded images for optimal results
- ›Use for batch processing to maximize throughput and cost savings
Quickstart
Pythonimport base64
import requests
response = requests.post(
'https://api.cohere.com/v2/parse',
headers={'Authorization': f'Bearer {API_KEY}'},
json={
'model': 'parse-v5.0',
'document': {'type': 'image_url', 'image_url': 'data:image/png;base64,{image}'},
'output_format': 'markdown'
}
)
markdown = response.json()['pages'][0]['markdown']['content']Pricing: $1.50 per 1,000 pages via API; $2,500-4,300/month via Model Vault
API model id: parse-v5.0
Benchmarks
| Benchmark | Score | Notes |
|---|---|---|
| ParseBench | 79.2 | Vendor-reported across three of five dimensions (tables, content faithfulness, semantic formatting); omits charts and visual grounding |
Source: Cohere Official Blog
Compare Cohere Parse 5
Compare Cohere Parse 5 with any other model
Build a comparison →All model comparisons →