Put DeepSeek on Your Own Computer
Company reimbursement policies, contracts, internal documents—you want AI to help you look them up, but you're not quite comfortable pasting them onto a web page.
The solution is to run the model on your own computer. Ollama's official documentation puts it plainly: when running locally, they cannot see your prompts or your data.
First, check how large a model your computer can run:
$ ollama psNAME SIZE PROCESSORdeepseek-r1:8b 5.7 GB 100% GPUdeepseek-r1:8b Can be embedded entirely into 8 GB of VRAM, runs the fastest.
What You'll Have When You Finish
A DeepSeek running on your computer, plus a knowledge base built from your own files: ask it "Can I get reimbursed for taxi rides on business trips," and it will find that exact line in your policy documents and answer you; if the documents don't say, it will say so.
Every command and every output in this course has been run on a real machine: Ubuntu, RTX 4080 SUPER 16 GB, Ollama 0.40.16.
This Course's Roadmap
| Stage | Lesson | What You'll Do |
|---|---|---|
| Getting Started | What is local deployment | Know what hardware you need and how it differs from the web version |
| Install Ollama | Run ollama -v in the terminal and print the version number | |
| Run DeepSeek-R1 | Chat with a local DeepSeek | |
| Put It to Work | Which model size to choose | Pick a size based on your VRAM |
| Call it via the API | Call the local model with curl and the OpenAI-compatible API | |
| FAQ | Where models are stored, how to move them, LAN access | |
| DeepSeek V4 Can it run locally | How much memory do V4 / V4.1 need, and why does Ollama only have the cloud version | |
| Knowledge Base | Build your own knowledge base | Ask questions against your own files |
Before You Start
- A computer: Windows 10 22H2 or later, or macOS Sonoma (14) or later, or Linux
- A dedicated GPU is best; you can still run small models without one, just slower
- Free up at least a dozen or so GB of disk space: Ollama itself plus an 8b model (5.2 GB)
Next Step
Start with What is local deployment.