The recently-released 3.3 version of my SketchUp AI Explorer (Experimental) extension now allows for an interesting and potentially (depending on your hardware) useful way to work with AI in SketchUp: You can now use local LLMs on your computer via a software like LM Studio as the AI provider for this extension.
This approach is 100% free and only depends on the hardware capabilities of your local computer. This means you don’t have to pay for tokens and you can even choose among a fairly large number of AI LLM models (including GPT-OSS, QWEN, GEMMA, DEEPSEEK,…). And nobody will ever train their AI systems on your prompts. One word of caution, though: This is largely an experimental approach and its effectiveness is very much dependent on your hardware (especially your GPU). But if you tweak this setup to your liking, it may open up new avenues.
How Does This Work?
To be able to do this, you will need two things:
- The Ai Explorer (Experimental) extension, which you can get from the Extension Warehouse or my website.
- LM Studio or a similar local LLM software. I like LM Studio because it is very user friendly but you can also use Ollama or others.
Once you have everything installed, you need to download a suitable AI model in LM Studio. Go to the Model Search button and browse their rather large list of models. You will need to weigh model size against your hardware capabilities, but you have enough choices. As a start, look for smaller versions of Google’s Gemma models or any of the NVIDIA ones. Once you have a model you like, download it in the software.
You are now ready to start LM Studio’s server. You can do that in the Developer tab. As you can see in the following image, clicking the toggle on the top left of the screen will do the trick. After this step, you will need to do two things, which I illustrate in the video tutorial:
- Load your downloaded LLM model using the button on the Developer tab. You may want to manually choose the model load parameters and adjust how much will be offloaded to the GPU (as much as possible, typically) and what the max. token size is (reduce if you are getting errors).
- Turn on authentication and get an API Key/Token. Click on the server settings button, switch authentication on and issue yourself a nice new token. Copy that key/token as you will need to paste it into the SketchUp extension.
Now you are ready to add the three main connecting pieces (all as text strings) to the AI Explorer extension in SketchUp. Open up its settings and paste these items:
- AI Model: Copy this from LM Studio’s page verbatim. There is a convenient copy button right at the top of the Developer tab under the server options.
- AI API Key: That is the key you just issued yourself in LM Studio.
- AI API Endpoint URL: You can copy this from LM Studio’s setup page once you start the server. Look for the chat completions endpoint there. By default, this is
http://localhost:1234/v1/chat/completions.
Once you have this all in place, you can just use the AI Explorer extension as usual, albeit now with a local model.
Some things to consider:
- Not every model can handle images/vision or files. If you want to upload screen shots or files, make sure you select the correct model that can handle these.
- You may want to google “optimize LM Studio”. This can give you tips for tweaking the software so that it works better with your local setup.
- Your computer may ventilate more than usual. This shows you nicely how much computing power is actually needed to run these models.
Video Tutorial
Did you find this useful? Let me know in the comments.