AI Web Center

// how to

How to Run DeepSeek in LM Studio Without the Command Line

Install LM Studio from lmstudio.ai, search for deepseek-r1 in its model browser, and download a version that fits your memory; the 8B version is 4.30 GB. Load it and chat in LM Studio’s window. Once the model is downloaded, LM Studio works entirely offline, and documents you chat with stay on your computer.

By Published Last updated 2 min read

The short version

LM Studio is the easiest way to run DeepSeek on your own computer if you’d rather click than type commands. It looks like a chat app, and after the download nothing you type leaves your machine.

Step by step

1. Check your computer

From LM Studio’s system requirements:

  • Mac: Apple Silicon (M1/M2/M3/M4), macOS 14 or newer, 16 GB of RAM or more recommended. 8 GB Macs may work with smaller models.
  • Windows: x64 with AVX2 support or ARM (Snapdragon X Elite), 16 GB of RAM recommended, at least 4 GB of dedicated graphics memory recommended.
  • Linux: x64 or ARM64, Ubuntu 20.04 or newer.

2. Install LM Studio

Download it from lmstudio.ai and install it like any other app.

3. Download a DeepSeek model

Search for deepseek-r1 in LM Studio’s model browser. The versions listed on LM Studio’s site:

Model Size
DeepSeek-R1-0528-Qwen3-8B 4.30 GB
DeepSeek-R1-Distill-Qwen-7B 4.30 GB
DeepSeek-R1-Distill-Llama-8B 4.50 GB
DeepSeek-R1-Distill-Qwen-14B 8.40 GB
DeepSeek-R1-Distill-Qwen-32B 18.70 GB
DeepSeek-R1-Distill-Llama-70B 40.30 GB

LM Studio says the smallest needs at least 4 GB of RAM and the largest may need up to 40 GB. Start with the 0528 Qwen3 8B version.

4. Load it and chat

Select the model, wait for it to load, and type. DeepSeek-R1 shows its reasoning before the answer.

5. Chat with your documents

Drag a document into the chat. LM Studio’s docs say the document stays on your machine and all processing happens locally.

Prompts to copy

DeepSeek’s R1 model card recommends putting your instructions in the message itself:

text
Read the attached document. List the five most important points,
then any dates, amounts or obligations, each with the sentence it comes from.
text
Rewrite this paragraph in plain English for a customer. Keep it under 80 words.
[paste]

Example

Setup: a Windows laptop with 16 GB of RAM and a 6 GB graphics card, running DeepSeek-R1-0528-Qwen3-8B.

Use: drafting replies to confidential HR emails offline. Answers take longer than the DeepSeek app and are less polished, but no text leaves the laptop.

Mistakes to avoid

  • Downloading the biggest model first. If it doesn’t fit in memory, it will be very slow or won’t load.
  • Expecting the DeepSeek app’s quality. These are distilled models built on Qwen and Llama, much smaller than the model behind chat.deepseek.com.
  • Forgetting the internet is still used for some things. Searching for and downloading models, runtimes and app updates need a connection.

Good to know

LM Studio can also act as a local server with OpenAI-style endpoints, so other apps on your computer or network can use your DeepSeek model. Requests stay local.

Related: run DeepSeek with Ollama and run DeepSeek locally.

Frequently asked questions

What computer do I need for LM Studio?

On a Mac, an Apple Silicon chip (M1 or newer) with macOS 14 or later, and 16 GB of RAM or more recommended. On Windows, a processor with AVX2 support, at least 16 GB of RAM and a graphics card with 4 GB of dedicated memory are recommended. Intel Macs aren't supported.

Is my data private in LM Studio?

LM Studio's docs say nothing you enter when chatting with downloaded models leaves your device, and documents you drag in for chatting are processed locally.

Which DeepSeek versions does LM Studio offer?

Its DeepSeek-R1 page lists distilled models from 7B to 70B, including DeepSeek-R1-0528-Qwen3-8B (4.30 GB) and DeepSeek-R1-Distill-Llama-70B (40.30 GB).

Sources

  1. LM Studio: deepseek-r1 models
  2. LM Studio Docs: system requirements
  3. LM Studio Docs: offline operation

// keep reading