DeepSeek-OCRon your own GPUs.

DeepSeek-OCR turns scans, PDFs and forms into text and structure on GPUs you control. The documents never leave your network.

Start the $4,900 pilot Compare all models

This card is the installation. Pick a size on its edge and it is re-issued.

LLM_Deploy

Card no. LD-131Issued 2026-10-27DeepSeek

DeepSeek-OCR

Pilot $4,900 fixed

Sizes: 3.3B MoE
Active params: ~0.57B
Context: 8K
Licence: MIT
Fits on: CPU only at 4-bit, any GPU
Endpoint: llm.yourcompany.internal/v1
Egress: 0
Owner: YOU

The 72-hour pilot with DeepSeek-OCR.

$4,900 fixed. DeepSeek-OCR is the most downloaded of the two and fits on CPU only at 4-bit. The others fit on the same hardware or less.

Live on your data

72 h$4,900 fixed

DeepSeek-OCR live on your data in 72 hours. The clock starts after access and prerequisites.

Your endpoint

/v1OpenAI-compatible

OpenAI-compatible endpoint. Your applications switch by changing the base URL.

Your hardware

DeepSeek-OCRCPU only at 4-bit, any GPU
DeepSeek-OCR 2CPU only at 4-bit, any GPU

The readout

Evalon your data

Reference evaluation on your data. Readout and production scoping.

We size the hardware and source it at cost, with no markup. Ballpark: one H100 80GB ≈ $25–30k; a 4×H100 node ≈ $180–240k; smaller models run on far less.

Same SDK. Change the base URL.

The model is one parameter. Everything else in your application stays as it is.

client = OpenAI(
    base_url="https://llm.yourcompany.internal/v1",   # inside your network
    api_key="your-key",
)
page = client.chat.completions.create(
    model="deepseek-ocr",
    messages=[{"role": "user", "content": [
        {"type": "image_url", "image_url": {"url": scan_data_url}},
        {"type": "text", "text": "Convert this page to Markdown."},
    ]}],
)

Licence, plainly.

DeepSeek-OCRMITCommercial use: permitted
DeepSeek-OCR 2Apache 2.0Commercial use: permitted

DeepSeek-OCR is released under the MIT licence, which permits commercial use, modification and private deployment.

DeepSeek-OCR 2 is released under Apache 2.0: commercial use, modification and private deployment are permitted.

We read the current licence text against your use when we scope the pilot.

Published by DeepSeek. We are not affiliated with the vendor. This is a summary, not legal advice.

Related models.

Sizing for every model and GPU

Talk to an engineer.

Tell us the model, the data and the GPUs you have. An engineer from Limassol or Auckland scopes the 72-hour pilot with you.

Start the $4,900 pilot Talk to an engineer

limassol.officeLocal time

Limassol HQ

Cyprus, 24/7

+357 25 123877

auckland.officeLocal time

Auckland

New Zealand, 24/7

+64 9 887 9088