[ Video ] Local AI: Testing Domyn Edge 4B on a 16GB MacBook Pro
For the 16th episode of my series in which I test local AI models that fit on a 16GB MacBook Pro, I chose Domyn Edge 4B. It’s a 4-billion-parameter model from Domyn (obviously), a Milan-based company that rebranded about
For the 16th episode of my series in which I test local AI models that fit on a 16GB MacBook Pro, I chose Domyn Edge 4B. It’s a 4-billion-parameter model from Domyn (obviously), a Milan-based company that rebranded about a year ago. The same company was selected by the European Commission to train one of Europe’s largest open-weight models (targeting up to 1 trillion parameters). The model I tested is the smaller, already-released piece.
It uses a custom architecture, and that means the usual tools I use for my tests wouldn’t load it. So I converted it myself to an MLX 4-bit quant. The resulting model is roughly 2.2–2.4 GB and runs comfortably on a 16 GB MacBook Pro with room left for other apps.
I tested it with mlx_lm.chat (not mlx-serve) and you can get the model yourself from huggingface if you want (link in the video description).
First prompt: “What is a dolphin?”, as usual.
The model started very fast, it showed its thinking process and produced a clear, correct answer in about 30 seconds. Overall, generation felt snappy for a 4B model on this hardware.
Second prompt: a small utility that counts words in files inside a folder.
After some internal reasoning it settled on a simple Python script that actually works, plus short usage instructions. It also briefly considered a bash version before choosing Python.
Overall the model is a pleasant surprise. Text generation is solid and the size makes it a very practical choice. Coding performance looks usable for small tasks; it’s also a good candidate for my longer coding series.
All links are in the video description.
As usual, screen recordings in the video are unedited and not sped up, so you get the exact experience of what it is to run it on consumer hardware.
If you like this kind of content, like, share and subscribe to my YouTube channel!
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.