LM Studio is a free desktop app for Windows, macOS and Linux that lets you download and run open AI models like Llama, Qwen, Gemma, DeepSeek and gpt-oss on your own computer, with a ChatGPT-style interface and a local OpenAI-compatible API.
Table of Contents
System Requirements
| Minimum | Recommended | |
|---|---|---|
| Windows/Linux | AVX2 CPU, 16 GB RAM | NVIDIA/AMD GPU with 8–24 GB VRAM |
| macOS | Apple Silicon (M1+) | M-series with 32 GB+ unified memory |
| Disk | 10 GB+ | SSD with space for models |
How to Run an LLM in LM Studio
- Download LM Studio from lmstudio.ai and install it.
- Open the Discover tab and search for a model (for example gemma 4, qwen or gpt-oss).
- Pick a quantization that fits your GPU (LM Studio marks which ones fit) and click Download.
- Go to Chat, load the model and start chatting.
Use LM Studio as a Local API
Open the Developer tab and start the server, or use the lms command-line tool:
lms server start
curl http://localhost:1234/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model":"gemma-4-12b","messages":[{"role":"user","content":"Hello"}]}'Which Model Fits Your Hardware?
| VRAM / memory | Model examples |
|---|---|
| 8 GB | Gemma 4 E4B, Llama 3.1 8B, Qwen 3.5 9B |
| 16 GB | Gemma 4 12B, gpt-oss-20b, Ministral 3 14B |
| 24–32 GB | Gemma 4 31B, Qwen 27B, Devstral Small 2 |
| 64 GB+ (Mac) | gpt-oss-120b, 70B models |
LM Studio vs Ollama vs Oobabooga
| LM Studio | Ollama | Oobabooga | |
|---|---|---|---|
| Interface | Desktop app | CLI + API | Web UI |
| Best for | Beginners on desktop | Servers and developers | Power users |
| Headless server | Limited | Excellent | Yes |
For a 24/7 AI server, Ollama on a GPU server is usually the better fit.
FAQ
Yes, LM Studio is free for personal and work use.
No, but a GPU (or Apple Silicon) makes it much faster.
It has a headless mode and the lms CLI, but Ollama or vLLM are more common for servers.
Conclusion
LM Studio is the friendliest way to run open AI models on your own computer. When you need a model available 24/7, move it to a GPU server from GPU Mart or HOSTKEY.