LM Studio Update: New Models, New Tools, Smarter Setup
In this LM Studio update, I walk through the key changes that matter most: setup, newer models, smarter load settings, useful plugins, and LM Link. This is a practical follow-up to my earlier LM Studio video for anyone who wants a simple, up-to-date overview.
Original LM Studio Video:
• Your First Local AI Model in Just 5 Minutes
Mentioned Models:
Nemotron (NVIDIA)
Qwen (Alibaba)
Mistral (Mistral)
Gemma (Google)
This is a longer walkthrough, so I included both a quick timeline and a more detailed jump list below.
Quick Timeline
0:00 Intro
1:25 Setup
4:07 Models and Loading
8:36 Load Settings
13:06 PDFs and Tokens
17:20 Web Search Plugins
21:10 LM Link
25:40 Final Thoughts
Detailed Timeline
0:00 Intro
0:19 What this LM Studio update covers
1:25 Installation and setup
2:27 Check your hardware and VRAM
3:09 CUDA vs CPU and runtime settings
4:07 Loading a model
4:49 Which models to try
7:08 Download and install a model
8:02 Model loading basics
8:36 Context length and GPU offload explained
10:20 Advanced settings in plain English
12:40 First prompt test
13:06 Working with PDFs in LM Studio
15:13 Large PDF test and token comparison
16:47 Do you really need more tokens?
17:20 How to add web search
18:13 Installing the web plugins
19:21 Using web access in chat
19:45 Managing and disabling plugins
20:39 RAG and chatting with files
21:10 LM Link overview
21:44 Requesting access and signing in
22:27 Connecting another device
23:16 Running a model remotely
24:49 Turning LM Link off
25:40 LM Studio pros and cons
27:45 My personal take
28:26 Outro
Links:
https://lmstudio.ai/danielsig (plugins)
https://developer.nvidia.com/nemotron (Nemotron)
https://huggingface.co/Qwen (Qwen)
https://docs.mistral.ai/getting-start... (Mistral)
https://deepmind.google/models/gemma/ (Gemma)
Sources:
https://www.nature.com/articles/s4159...