Run Private GenAI on Your Local Machine with LM Studio
- Jeenie Yoon
- 3 hours ago
- 3 min read
Generative artificial intelligence (GenAI) has many potential uses in higher education. However, cloud-based AI products like ChatGPT, Gemini, Claude, and Perplexity raise data privacy concerns. In this post I write about LM Studio, an application that allows users to download and run GenAI models locally without sending data to companies like OpenAI, Google, or Microsoft.
LM Studio: Features and Functionality
LM Studio is a free desktop application designed to facilitate access to large language models (LLMs) (LM Studio, n.d.). It is compatible with macOS, Windows, and Linux, and allows users to search for, download, and execute various LLMs directly on their local machines (LM Studio, n.d.). LM Studio seamlessly connects with leading platforms for AI models, such as Hugging Face (LM Studio, n.d.). This provides users with straightforward access to a wide variety of pre-built models, which can be run efficiently using GPU acceleration. Essentially, this integration allows you to explore the latest AI models without the complex process of manual setup and troubleshooting.
LM Studio features an interface similar to that of ChatGPT; however, at the time of this writing, the file upload functionality is restricted to Word, PDF, and plain text documents (LM Studio, n.d.). Beyond basic chat functionality, LM Studio offers some customization methods. Users can adjust parameters like “Temperature” to refine model output, define system prompts to guide AI behavior, and utilize different runtimes to improve performance (LM Studio, n.d.). For users with more technical experience, LM Studio includes a developer mode with advanced functionalities such as server configuration and API endpoints (LM Studio, n.d.). This API capability empowers advanced model customization for downstream tasks and integrations with third-party applications.
Addressing Privacy Concerns through Local GenAI
The primary advantage of a local AI platform like LM Studio in education is enhanced privacy. Unlike cloud-based services that transmit data to remote servers, LM Studio ensures that interactions, including prompts, conversation histories, and uploaded documents, remain on the user's device (LM Studio, n.d.). That said, please note that LM Studio still collects some basic information when you search for, download, or update models. This information is strictly limited to what's needed for the app to work ("need-to-know") and isn't linked to you personally. A table detailing when and what LM Studio collects from you is in the Appendix below.
For individuals working with sensitive materials, LM Studio’s local execution significantly reduces the risks tied to external servers. This is particularly beneficial for faculty handling sensitive data, as it ensures their information isn't exposed to third-party servers (LM Studio, n.d.). In contrast, cloud-based AI services inherently introduce vulnerabilities due to remote data storage, heightening the risk of breaches and unauthorized access (Baig & Malik, 2024). This makes a local solution like LM Studio a compelling choice for privacy-conscious educators.
Install LM Studio on Your Computer
Here are two guides that walk you through how to install LM Studio on macOS and Windows, respectively.
macOS
Step 1: Visit the application download site from LM Studio
Step 2: Follow this screen recording to download and install LM Studio – LM Studio Installation - macOS.mov
Windows
Guide from Randy Hanley.
Technical Requirements for Running LM Studio
Faculty will need to consider their computer's technical specifications to ensure optimal performance with LM Studio (LM Studio, n.d.). The application is designed to run on various systems, but the size and complexity of AI models will affect resource requirements. Below is a table that lists the spec recommended by the LM Studio team at the time of writing (LM Studio, n.d.).
Recommended System Specifications for LM Studio
Specification | Minimum | Recommended | Optimal (for larger models) |
Operating System | macOS 13.4+, Windows 10/11, Linux Ubuntu 20.04+ (Beta) | macOS 14.0+, Windows 10/11, Linux Ubuntu 20.04+ (Beta) | macOS 14.0+, Windows 10/11, Linux Ubuntu 20.04+ (Beta) |
CPU | AVX2 Support (x64), Apple Silicon (Mac) | Multi-core Processor (AVX2 or Apple Silicon) | Multi-core Processor (AVX2 or Apple Silicon) |
RAM | 8GB (running smaller models) | 16GB | 32GB+ |
GPU (VRAM) | N/A (Mac CPU), 4GB (Win/Linux) | 4GB+ (Win/Linux), Apple Silicon (Mac MLX) | 8GB+ (Win/Linux), Apple Silicon (Mac MLX) |
Disk Space | 20GB Free | 50GB+ Free (SSD Recommended) | 50GB+ Free (SSD Recommended) |
Please note that the system requirements are constantly changing so check the LM Studio’s site for the most up-to-date information
Appendix
When and What LM Studio Collects from You (LM Studio, n.d.).
Context of Data Collection | Specific Information Collected | Purpose of Collection | Notes |
Checking for App Updates | Device and App Information (version, OS, build) | Deliver appropriate app updates; understand the user base's software and hardware spread. | Includes IP address (via CDN); does not collect personal identifiers or track individual behavior. |
Searching for and Downloading Models | Anonymized Search Queries and Download Information | Help users find models; facilitate downloads; determine model popularity and frequency. | Anonymized; not tied to individual users; no individual search patterns tracked or user profiles created. |
Direct Email Communication | Email Address and Content of Email | Respond to user questions; provide support. | |
User Activity (Chats & Documents) | No data collected or transmitted from your system. | User messages, chat histories, and documents are processed and saved locally by default. | LM Studio cannot see your chats or documents. This is a fundamental privacy feature highlighted in the policy. |
References
Baig, A. & Malik, O.I. (2024). Generative AI Privacy: Issues, Challenges & How to Protect? - Securiti. https://securiti.ai/generative-ai-privacy/
LM Studio. (n.d.). Discover, download, and run local LLMs. Retrieved April 22, 2025, from https://lmstudio.ai/
LM Studio. (n.d.). Download an LLM | LM Studio Docs. Retrieved April 30, 2025, from https://lmstudio.ai/docs/app/basics/download-model
LM Studio. (n.d.). Chat with Documents | LM Studio Docs. Retrieved April 30, 2025, from https://lmstudio.ai/docs/app/basics/rag
LM Studio. (n.d.). Config Presets | LM Studio Docs. Retrieved April 30, 2025, from https://lmstudio.ai/docs/app/presets
LM Studio. (n.d.). LM Studio as a Local LLM API Server | LM Studio Docs. Retrieved May 2, 2025, from https://lmstudio.ai/docs/app/api
LM Studio. (n.d.). LM Studio Privacy Policy. Retrieved May 2, 2025, from https://lmstudio.ai/app-privacy
LM Studio. (n.d.). System Requirements | LM Studio Docs. Retrieved May 6, 2025, from https://lmstudio.ai/docs/app/system-requirements