Parse 5
#91 in Frontier LLMsCohere · v5 · 3× · last seen Aug 31, 2026
25
Momentum
Parse 5 (Model ID: parse-v5.0) is a Vision-Language Model for document processing released by Cohere on August 27, 2026. It converts PDFs, presentation slides, and images into structured Markdown, including tables, lists, form fields, and Bounding-Boxes for visual elements, without a separate OCR step. The model has 2.3 billion parameters, is based on the Cohere Labs architecture "North-Micro-Vision-Instruct," and is designed for price-performance ratio rather than peak accuracy. It is available via the Cohere Parse API, Microsoft Foundry, AWS SageMaker, and in the single-tenant model Vault.
Momentum trend
02.06.31.08.
Features
| Key Benchmark (%) | ParseBench: 79.2 (average across tables, content faithfulness, semantic formatting) |
| Context Window (Tokens) | 8,192 tokens |
| Multimodality | Vision-language model: processes PDF, PPT and JPEG pages (image input) and returns Markdown/HTML output with text, tables, image descriptions and bounding boxes |
| Platform | Cohere Parse API, Microsoft Foundry, AWS SageMaker, Model Vault (single-tenant) |
| Price per 1M Tokens | $1.50 per 1,000 pages (page-based API pricing, not standard per-token rate) |
| Release Date | August 27, 2026 |