🧠 Choosing a Local Inference Server

To run a local LLM with Elite Intel, you need an inference server - software that loads the AI model and serves it over a local API Elite Intel can talk to. Think of it like a local version of the cloud AI services, running entirely on your own hardware.

Elite Intel supports two of them: Ollama and LM Studio. Both work. Both use the same models. Pick one and follow its guide - you can always switch later in settings.

Can my GPU pull this off?

Here is a good reference table provided by Kevin Rank GPU Reference guide

Which one do you use?

I use LM Studio with matrixportalx/Tulu-3.1-8B-SuperNova-Q4_K_M-GGUF This one is fast. Very fast! The same model on Ollama runs noticeably slower. The app is tuned to work with this model. All other models use at your own risk. It may work fine, it may not. Let me know.

Why tulu3.1:8b Supernova specifically?

This explains it

---

Install Guides

Inference Server
✅ LM Studio - Linux	Fast, more model flexibility - guide shows how to setup as a server
✅ LM Studio - Windows	Fast, more model flexibility - got GUI
Ollama - Linux	Recommended if you have the hardware to run it
Ollama - Windows	Recommended if you have the hardware to run it

---

Ollama vs. LM Studio at a Glance

	Ollama	LM Studio
Speed	Slower	Faster
Prefered model	tulu3:8b	tulu-3.1-8b-supernova (Q4_K_M Variant)
Best for	Set it and forget it	More control over what you load
Install	One script, done	One script, done
Runs as	System service (auto-starts on boot)	Manual start, or opt-in auto-start
Model tuning	Modelfile baked into the model	Flags at load time
Windows auto-start	✅ Works out of the box	Requires desktop app or Task Scheduler
Linux auto-start	✅ systemd service included	Manual systemd setup
Model source	Ollama library	HuggingFace (GGUF)
API port	11434	1234
GUI	None (CLI only)	Optional desktop app

---

Just Tell Me Which One to Pick

Pick Ollama if:

• You want something that just works after a five-minute install and never think about it again
• You're on Windows and don't want to mess with startup configuration
• You're new to local LLMs

Pick LM Studio if:

• You want a desktop GUI to browse, download, and manage models
• You're already familiar with HuggingFace and GGUF model files
• You want to experiment with different models without writing Modelfiles
• You're running a dedicated AI box on your network and want a clean headless server

Either works fine if:

• You have an NVIDIA RTX 3090 24GB (equivalent or better). It is not the speed, it is the VRAM. RTX 5090 with 12GB VRAM will not do.
• You're running Elite Dangerous and the LLM on the same machine
• You want to point Elite Intel at a separate PC on your network

---

Community 👉Matrix👈