RT if you want day-zero support for OpenAI new open-weights model to run fast on your laptop with MLX.

Been really fun watching Marko bio/acc discover he's got an incredible talent for voxel art these past two weeks - this microscope is AWESOME and not even the coolest of what he's made so far 🤯 V. stoked to share more of the project once it's cooked a lil longer hahaha

For reference, Gemma-27b runs at about ~10tps on my laptop, versus Qwen-30b at ~60tps To put it another way: in full power mode I can run about 350 prompts with Gemma-27b before my laptop runs out of juice on a single charge. Qwen will manage about 2,000 🥳

Made a fun little user interface for running these thinking models in my terminal! Had a lot of fun designing the intro animation hehe

The war on science in the US is already affecting private sector research like AlphaFold. Bears repeating but the private sector builds on top of things created by academic research for the public good. This hurts everyone.

The MLX project is by far and away one of the coolest things going on at Apple right now, and is led by some really clearly talented people. Been saying this for over a year. Which just makes me so furious that this rumour may well be true! What the actual f*ck, Apple