Teams feeding large document volumes into search or RAG pipelines get a cheap parser at $1.50 per 1,000 pages, with a free evaluation key.
Kenning's take
What it does
Cohere Parse is a small vision language model that reads business documents, including tables, forms and images, and returns structured Markdown with bounding boxes. It is meant for high-volume jobs such as preparing documents for search, RAG pipelines and AI agents. It is available through the Cohere API, in Cohere Compass, or as a private deployment.
- Extracts text, tables, lists, forms and image captions
- Markdown output with tables as HTML and bounding boxes
- Stable in nine languages including English, French and German
- Accepts PDF, PPT and JPEG
- 2.3B parameters, can run on private infrastructure
Who it suits. Developers and enterprise teams processing large volumes of documents for search and AI retrieval.
Key facts
- Category
- AI models
- Made by
- Cohere
- Based in
- Canada
- Released
- 27 August 2026
- How to get it
- Cohere API (free evaluation key with rate limits, paid production key), Cohere Compass, Model Vault or private deployment; free demo on Hugging Face Spaces
- Works on
- API
- Free plan
- Free evaluation API key with usage limits, and a free Hugging Face Space demo
- API price
- $1.50 per 1,000 pages
Pricing
| Plan | Price | Billed |
|---|---|---|
| Model Vault Parse 5 Medium | $4 | hour per instance ($2,500 a month) |
| Model Vault Parse 5 XL | $7 | hour per instance ($4,300 a month) |
FAQ
When was Cohere Parse released?
Cohere announced Parse on 27 August 2026. The current version is parse-v5.0.
How much does Cohere Parse cost?
Through the Cohere API it costs $1.50 per 1,000 pages. A free evaluation key and a Hugging Face demo are available for testing.
What file types does Cohere Parse accept?
PDF, PPT and JPEG, and it returns Markdown.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.