How to Run DeepSeek Locally on Your PC in 2026 (Ollama & Open WebUI Guide)

With the release of DeepSeek-R1 and DeepSeek-V3, open-weights artificial intelligence matched the reasoning and coding capabilities of closed proprietary giants like OpenAI and Google.

However, the true power of DeepSeek lies in its open licensing (MIT): you can download the model weights and run them entirely on your own hardware. No monthly subscription, no rate limits, and 100% data confidentiality.

Here is the complete step-by-step tutorial to running DeepSeek on Windows, macOS, or Linux in 2026 using Ollama and Open WebUI.

---

Hardware Sizing Guide: Which DeepSeek Model Can You Run?

DeepSeek-R1 was released in several quantized and distilled sizes to accommodate everything from budget laptops to high-end workstations:

Model Version	Parameter Count	Minimum RAM / VRAM	Recommended Hardware	Tokens/Sec Speed
DeepSeek-R1: 1.5B	1.5 Billion	4 GB RAM	Any modern laptop / CPU	45+ tok/s
DeepSeek-R1: 7B / 8B	7–8 Billion	8 GB VRAM / 16 GB RAM	RTX 3060 / Apple M1 16GB	25–35 tok/s
DeepSeek-R1: 14B	14 Billion	12 GB VRAM / 24 GB RAM	RTX 4070 / Apple M2 24GB	20–30 tok/s
DeepSeek-R1: 32B	32 Billion	20 GB VRAM / 32 GB RAM	RTX 3090 / 4090 / Apple M3 36GB	15–22 tok/s
DeepSeek-R1: 70B	70 Billion	48 GB VRAM / 64 GB Unified	2x RTX 3090 or Mac Studio 64GB	8–15 tok/s

---

Step 1: Install Ollama (The Inference Engine)

Ollama is the premier local model runner. It handles model quantization, memory offloading to GPUs, and serves a local REST API.

1. Go to ollama.com and download the installer for your OS (Windows, macOS, or Linux).
2. Run the installer. On Windows and Mac, Ollama will run quietly in your system tray.
3. Verify the installation by opening your terminal or PowerShell and typing: ``bash ollama --version ``

---

Step 2: Download and Run DeepSeek in Your Terminal

To run your first prompt, choose the model that fits your hardware:

For Standard Laptops (8GB–16GB RAM):

For Gaming PCs / MacBooks (16GB–32GB RAM):

For High-End Workstations (32GB+ RAM / RTX 4090):

Ollama will automatically download the quantized GGUF weights. Once complete, you will see an interactive prompt where you can chat with DeepSeek immediately in your terminal!

---

Step 3: Install Open WebUI (The ChatGPT-Style Graphical Interface)

While the terminal is great for quick tests, a visual web interface with chat history, document uploads (RAG), and model switching makes the experience identical to ChatGPT.

The cleanest method to install Open WebUI is via Docker:

1. Install Docker Desktop:

Download and install Docker Desktop from docker.com.

2. Run the Open WebUI Container:

Open your terminal and run the following single command:

(Note: On Mac/Linux, this command connects Docker directly to your local Ollama port 11434.)

3. Open the Interface:

1. Open your web browser and go to: http://localhost:3000
2. Create an admin account (stored locally on your hard drive).
3. In the top model dropdown, select your downloaded deepseek-r1 model.
4. You now have a private, lightning-fast ChatGPT clone running 100% locally on your machine!

---

Power User Tips: Activating Reasoning Chain in WebUI

• View Thinking Process: DeepSeek-R1 emits tokens. Open WebUI automatically renders these inside an interactive collapsible accordion, allowing you to inspect the model's inner reasoning chain before reading the final answer.
• Document Analysis (RAG): Click the + icon in the chat bar to upload private PDF documents, financial spreadsheets, or source code. DeepSeek will parse and answer questions about them without sending a single byte to external servers.