72 hours, four big open weights drops. The only one that fits in 24GB came from NVIDIA, of all companies.

Explain Like I'm Normal

A massive week for open weights saw Meta release Muse Spark Open, a massive 816B parameter model that technically runs on 'desktops' only if you spend $38,000 on hardware. Surprisingly, NVIDIA stole the spotlight with Nemotron 3.5 Bolt, a 74B model optimized via quantization to fit within 24GB of VRAM. This makes high-end agentic performance accessible to standard consumer GPUs like the RTX 3090/4090 for the first time.