If you are like me, are you tinkering around with self-hosted small language models and running these locally? I noticed that Microsoft open-sourced their Phi-4 language model. You can get this up and running fairly quickly with LM Studio.
Just in case you didn't know LM Studio is a free tool that provides a really nice GUI to work with local language models and allow you to easily download these and run them on your own hardware.
You can download LM studio here: LM Studio - Discover, download, and run local LLMs
Once you have it downloaded, you can search for new models and when I pulled it up, the phi-4 model was actually listed at the top of the list (probably because it is new). However, if you don't see it there, you can search for it.
It is a little hefty at 9 GB download, but it is still amazing to me the power of these models even at 9 GB in size, that is not that large for the training data you get. Once you get the model downloaded, you will see the button to the side to Load Model.
This will automatically load the model into LM Studio. You can still do this if you don't do this here. When you go to your Chat screen, it will give you the option to load a model, where you can select the phi-4 model. Once the model is loaded, you can simply navigate to your chat screen on the left that looks like a "text" icon and start chatting (if you have already loaded the phi-4 model), otherwise, load the model and then start your chat! So far it looks pretty cool and powerful.
You can take a look at the hugging face download here: