Elevator Pitch

• A 2.5-year-old MacBook Pro can now run the powerful GLM-4.5 Air coding model locally, generating functional JavaScript games and graphics, demonstrating how far local LLMs have advanced.

Key Takeaways

• The 44GB 3bit quantized GLM-4.5 Air model runs efficiently on a consumer 64GB MacBook Pro, producing working Space Invaders code on the first try.
• The process involves using the mlx-lm library and some technical setup, but the model delivers impressive coding results and even generates SVG graphics.
• Local coding models have rapidly improved, with 2025 models like GLM-4.5 Air rivaling cloud-based solutions in capability.

Most Memorable Aspects

• A laptop from early 2023 can now run a 106B parameter model and generate a playable game with no edits needed.
• The model generated both a working game and a whimsical SVG image of a pelican riding a bicycle.
• Running these models requires most available RAM but is now feasible for individuals with high-end laptops.

Direct Quotes

• "I still think it’s noteworthy that a model running on my 2.5 year old laptop... is able to produce code like this—especially code that worked first time with no further edits needed."
• "It’s interesting how almost every model released in 2025 has specifically targeting coding. That focus has clearly been paying off: these coding models are getting really good now."
• "Two years ago when I first tried LLaMA I never dreamed that the same laptop I was using then would one day be able to run models with capabilities as strong as what I’m seeing from GLM 4.5 Air..."