LM Studio step by step: run your first local LLM
Install LM Studio, pick a model that fits your GPU or Mac, load it and chat — plus the local server. A beginner walkthrough.
LM Studio is a free app for Windows, macOS and Linux that downloads and runs models on your machine, with no command line. Interface labels may differ between versions.
1. Check what your machine supports
Before downloading anything, enter your graphics card or Mac in the configurator: it lists the models that fit and the memory they need. Simple rule: the model, its quantization and the context must fit in VRAM (or unified memory on a Mac).
2. Install LM Studio
Download the installer from the official site, lmstudio.ai, and run it.
3. Pick and download a model
In the model search, look for a model name (for example Qwen3 or Llama). Choose a GGUF file; the Q4_K_M quantization is the best size/quality trade-off (see Q4 vs Q5 vs Q8). Here is the memory they need:
| Model | Q8_0 | Q4_K_M |
|---|---|---|
| Llama 3.1 8B Instruct | ~10.2 GB | ~6.7 GB |
| Qwen3 8B | ~10.5 GB | ~6.9 GB |
| Qwen3 14B | ~17.5 GB | ~11 GB |
| Gemma 3 27B IT | ~30.6 GB | ~18.6 GB |
File sizes: not available until imported from the model repository. Memory = weights (estimated from bits per weight) + KV cache + runtime overhead + safety margin.
4. Load the model and chat
Open the chat tab, select the downloaded model and load it. If LM Studio offers a GPU offload setting, set it to the maximum when the model fits in VRAM; otherwise lower it, or pick a smaller model. A longer context length uses more memory.
5. Use it from your apps
LM Studio can start a local server compatible with the OpenAI API, which other tools can call by pointing to your machine. Everything stays local.
Going further: local LLMs for beginners and how much VRAM.
EstimateData updated 38 days agoSource: Run LLM Locally engine