How to install and use DeepSeek on your PC

February 1, 2025

In a previous post, we explored how to install LM Studio on a standard personal computer to run locally LLMs (more or less modest) without relying on an internet connection or cloud-based services, as you need when using ChatGPT.

Language models and LLMs have democratized access to advanced AI, making it more accessible to individual users and small businesses. These models are becoming increasingly capable of generating coherent and creative text, writing code, analyzing data, and, in some cases, generating visual content using specialized models.

Returning to LM Studio and similar software, these programs allow users to run language models on relatively modest personal computers if they meet certain requirements, such as a modern processor and sufficient RAM—ideally 16 GB or more.

■ This transforms these language models into personal AI assistants that help with tasks such as translation, text summarization, code review, or content creation.

What is DeepSeek R1 Distill

The startup DeepSeek has developed advanced language models like R1, V3 o Janus, which stand out for their efficiency and performance with less training resources than similar models.

These models are open source and non-proprietary. This means that you could download DeepSeek R1 and install it on your computer, assuming it meets the hardware requirements.

It also means that you can create distilled versions (distill) of the R1 model, smaller and with lower hardware requirements. You can run these versions on a regular PC (and computers Copilot+) allowing you to run locally AI applications efficiently.

How to distill an LLM model?

Distilled models are smaller, more efficient versions of large language models created through a process known as distillation. This method reduces the model's size and resource consumption while preserving much of its key knowledge.

The distillation process works by having a large, complex model to train a smaller model that learns to replicate its behavior using significantly lower computational demands. As a result, distilled models can run on consumer-grade computers without requiring specialized hardware.

However, distillation has the main drawback of losing certain details, specialized knowledge, and complex reasoning abilities. This can impact their accuracy or capacity to handle sophisticated or domain-specific queries.

Installing DeepSeek R1 Distill on your computer with LM Studio

Once you have LM Studio (or a similar program) installed and updated to the latest version, follow these steps:

Open LM Studio and access the model catalog by clicking the magnifying glass icon.

Look for the model by typing 'DeepSeek R1' into the search bar.
—The model DeepSeek R1 Distill (Qwen 7B) is well-suited to a modest personal computer.

Select the model and click Download.
Once the download is complete, switch to the Chat tab in the sidebar.

In the Select a model to chat with dropdown menu, choose DeepSeek R1 Distill (Qwen 7B) or any other DeepSeek model you have downloaded.
Start a conversation with DeepSeek in the Chat window.

Although, as we have mentioned, distillation inevitably leads to some loss of detail and reasoning ability, having access to advanced models like DeepSeek R1 Distill on personal computers enables users to run powerful models in more diverse settings.

Furthermore, it allows individuals to gain a deeper understanding of how these models work, explore their limitations, and experiment with their adaptation and customization. This opens new possibilities for personal and professional applications, including education, learning, and research.

■ DeepSeek in Perplexity. If you are a Perplexity Pro user you can use the full version of DeepSeek by selecting 'Reasoning with R1' from the available models.

—You can use it for a variety of applications, such as data analysis, summarize, translate and text generation or programming assistance.