Serve LLMs with LM Studio

Basic idea

LM Studio is an application for discovering, running, and serving large language models entirely on your own hardware. You can run local LLMs like gpt-oss, Qwen3, Gemma3, DeepSeek, and many more models privately and for free.

This playbook shows you how to deploy LM Studio on your hardware platform to run LLMs locally with GPU acceleration, so the machine acts as your own private, high-performance LLM server.

LM Link (optional) lets you use models on your hardware platform from another machine as if they were local. You link the hardware platform and your laptop (or other devices) over an end-to-end encrypted connection, so you can load and run models remotely without being on the same LAN or opening network access. See LM Link and Step 4 in the Instructions tab.

What you'll accomplish

You'll deploy LM Studio on your hardware platform to run Nemotron 3.5 Super VL (nvidia/nemotron-3.5-super-vl), and use the model from your laptop. More specifically, you will:

Install llmster, a headless, terminal-native LM Studio on the hardware platform
Run LLM inference locally via API
Interact with models from your laptop using the LM Studio SDK
Optionally use LM Link to connect the hardware platform and laptop over an encrypted link so remote models appear as local (no same-network or bind setup required)

What to know before starting

Required:

Working with terminal / command line interfaces
Understanding of REST API concepts
Terminal or SSH access to your hardware platform

Optional:

Familiarity with local network access / remote SSH to your hardware platform

Supported hardware platforms

Use the matrix below to confirm your hardware platform, recommended defaults, and whether multi-node applies.

Hardware platform	OS	Memory	Recommended default local settings	Multi-node capable hardware
DGX Spark	DGX OS	128 GB Unified Memory	llmster (`lms`) · API on port `1234` · example model `nvidia/nemotron-3.5-super-vl`	—

NOTE

Only platforms listed in the Supported hardware platforms table above are covered by this playbook.

Prerequisites

Hardware requirements

Supported hardware platform — see Supported hardware platforms matrix above
Minimum 65 GB memory for model inference (70 GB or above recommended for the example model)
At least 65 GB available storage for model downloads (70 GB or above recommended)

Software requirements

Client device (Mac, Windows, or Linux) to call the API or use LM Link
For LAN access without LM Link: client device and hardware platform on the same local network
Network access to download packages and models

Find model recipes

Browse supported models and download paths on the LM Studio model catalog. Use the catalog to explore other models beyond the example in this playbook.

Hardware platform	More recipes
DGX Spark	LM Studio model catalog

Nemotron 3.5 Super VL (nvidia/nemotron-3.5-super-vl) is the recommended model for this playbook. The Instructions tab uses it throughout.

Example models:

Model	Model path
Nemotron 3.5 Super VL (recommended)	`nvidia/nemotron-3.5-super-vl`
GPT-OSS-120B	`openai/gpt-oss-120b`

Use the Instructions tab for the base workflow.

LM Link (optional)

LM Link lets you use your local models remotely. You link machines (for example, your hardware platform and your laptop), then load models on the hardware platform and use them from the laptop as if they were local.

End-to-end encrypted — Built on Tailscale mesh VPNs; devices are not exposed to the public internet.
Works with the local server — Any tool that connects to LM Studio’s local API (for example, localhost:1234) can use models from your Link, including Codex, Claude Code, OpenCode, and the LM Studio SDK.
Preview — Free for up to 2 users, 5 devices each (10 devices total). Create your Link at lmstudio.ai/link.

If you use LM Link, you can skip binding the server to 0.0.0.0 and using the hardware platform’s IP; once devices are linked, point your laptop at localhost:1234 and remote models appear in the model loader.

Ancillary files

Sample scripts for Step 7 of Instructions (hosted by LM Studio docs):

run.js — JavaScript script for sending a test prompt
run.py — Python script for sending a test prompt
run.sh — Bash script for sending a test prompt

Time & risk

Estimated time: 30 MIN (including model download time, which may vary depending on your internet connection and the model size)
Risk level: Low
- Large model downloads may take significant time depending on network speed
Rollback: Downloaded models can be removed manually from the models directory. Uninstall LM Studio or llmster (see Cleanup in the Instructions tab).
Last Updated: 10/21/2026
- Hardware platforms matrix and Find model recipes for serving LLMs with LM Studio
- Named Nemotron 3.5 Super VL as the recommended model and removed Qwen3.6-35B-A3B from the example models