How to Install DeepSeek on Windows for Free (Run a $0 AI Assistant Locally)
Paid AI assistants like ChatGPT Plus run about $20/month — roughly $240 a year. DeepSeek gives you a comparable chat assistant for free, and while it has no official Windows installer, you can put it on your desktop in under a minute or run it fully offline. Here’s exactly how to do both, and how to pick the setup that saves you the most.
💡 The money angle: two legitimate, zero-cost setups can replace a paid AI subscription — one online in ~1 minute, one fully private and offline.
What DeepSeek Is and Why It’s Worth Setting Up
DeepSeek is a free AI chat assistant — similar to ChatGPT — that can write, summarize, code, and answer questions. Its R1 reasoning model competes with paid tools that charge a monthly subscription, which is the money angle here: for a lot of everyday tasks, a free DeepSeek desktop setup can replace a paid AI plan and save you the annual fee entirely.
There is no .exe DeepSeek app for Windows. Anyone selling or “installing” one for you is repackaging something free. You have two legitimate, zero-cost paths:
Turns the official website into a real desktop app in about a minute. Fastest, and free forever.
Runs DeepSeek’s models locally on your own hardware — private, no account, no internet needed after download.
Both are free. Choose the PWA for speed, or Ollama if you care about privacy — a point worth taking seriously given DeepSeek’s reported exposed database of user data in early 2025.
Method 1: Install DeepSeek as a Desktop App (PWA via Edge or Chrome)
A Progressive Web App (PWA) wraps the DeepSeek website in its own window with a Start Menu shortcut — it looks and launches like a native app. Works on Windows 10 and 11.
Using Microsoft Edge
- Open Edge and go to chat.deepseek.com.
- Sign in (or create a free account).
- Click the
···menu (top right) → Apps → Install this site as an app. - Name it “DeepSeek” and click Install.
- Windows creates a desktop and Start Menu shortcut. Done.
Using Google Chrome
- Open Chrome and go to chat.deepseek.com.
- Click the install icon (a monitor with a down-arrow) on the right of the address bar. No icon? Use the
⋮menu → Cast, save, and share → Install page as app. - Confirm Install.
⚠️ You now have DeepSeek in its own window, pinnable to the taskbar. It still runs online through the browser engine, so your chats go to DeepSeek’s servers — fine for casual use, but not private.
Method 2: Run DeepSeek Locally on Windows with Ollama
Want DeepSeek fully offline, private, and free of any subscription? Ollama is a free tool that downloads and runs open AI models on your own PC. Nothing leaves your machine.
- Download Ollama for Windows from the official Ollama site and run the installer.
- Open Command Prompt or PowerShell (press
Win + R, typecmd, Enter). - Pull and run a DeepSeek model with one command:
ollama run deepseek-r1:8b
- Ollama downloads the model (a few GB), then drops you into a chat prompt right in the terminal. Type your question and press Enter.
- To exit, type
/bye. To reopen later, run the same command — it won’t re-download.
🖥️ Want a nicer interface than a terminal? Install Open WebUI, a free browser-based chat window for your local models. If you have Docker installed, one command gets it running, then you chat with DeepSeek at localhost:3000 — same clean look as the website, but 100% local.
Once it’s set up, local DeepSeek costs nothing per month and works with no internet — no account, no data sent anywhere.
Which Model Size Should You Download?
DeepSeek R1 comes in several sizes. Bigger models are smarter but need more memory. Match the model to your PC so it actually runs well.
✅ Rule of thumb: with no dedicated graphics card, stick to 1.5b or 7b — they run on the CPU, just slower. With a modern GPU (8 GB+ VRAM), 8b or 14b feel snappy. The 32b model really wants a strong GPU or it will crawl. Start small; you can always pull a bigger model later.
Watch Out: Unofficial “Downloads” to Avoid
Because there’s no official installer, scammy pages fill the gap. Skip these:
- 🚫 Random GitHub “DeepSeek Windows app” wrappers. Most are just an Electron shell around the website — the same thing the free PWA does, but from an unknown developer who could inject anything. No upside.
- 🚫 Itch.io or software-download sites offering a “DeepSeek installer.” DeepSeek doesn’t distribute a Windows binary anywhere. These are repackaged files at best, malware at worst.
- 🚫 “DeepSeek Pro” paid apps. You never need to pay for a wrapper around a free service. That’s the whole money point of this guide — don’t hand cash to a middleman.
Stick to two sources only: chat.deepseek.com for the PWA and ollama.com for the local install. Everything else is a risk.
Cheaper and Better Alternatives to Consider
If DeepSeek isn’t a fit — or you want a backup that costs nothing — a few free options run the same way:
- Other Ollama models. The same
ollama runcommand works with Llama 3, Mistral, and Qwen. If DeepSeek feels slow on your hardware, a smaller Llama model may run better for free. - Free web assistants as PWAs. You can install ChatGPT’s free tier or Google Gemini as a desktop PWA using the exact Method 1 steps — a solid way to get a free AI assistant without any monthly bill.
- Browser AI for one-off tasks. If you only need occasional help, the plain website in a pinned tab costs nothing and skips setup entirely.
Between free local models and free web tiers, most people never need a paid AI subscription at all.
FAQ
Can I use DeepSeek on Windows completely offline?
Yes — with the Ollama method. After the model downloads, it runs entirely on your PC with no internet. The PWA method still needs a connection.
Is DeepSeek private if I install it locally?
Yes. Local models via Ollama send nothing to DeepSeek’s servers — everything stays on your machine. This matters because of the reported 2025 data exposure affecting the online service.
Can I run DeepSeek without a graphics card (GPU)?
Yes. The 1.5b and 7b models run on your CPU and normal RAM. They’re slower than on a GPU but work fine for everyday questions on a typical laptop.
Do I need an API key or a paid plan?
No. The PWA uses the free chat account, and Ollama needs no key or payment at all. An API key is only for developers building apps — not for chatting.
How is the local version different from the website?
The website always runs the newest full-size model but sends your chats to DeepSeek’s servers. The local version is private and free forever, but runs a smaller model limited by your hardware.
Bottom line: use the PWA for a one-minute free desktop app, or Ollama for a private, offline assistant — either way you get a capable AI without paying a monthly subscription.
KO 
EN