Open-Weight Model Surge: Gemma 4, GLM-5.1 (1.5TB), and Qwen3.6-35B Now Run on Laptops

simonwillison.net·medium signal

Simon Willison's PyCon talk highlights a dramatic open-weight model surge: Google's Gemma 4 (April) is their most capable open model yet, China's GLM-5.1 is a 1.5TB open-weight release, and Qwen3.6-35B-A3B fits in 20.9GB — running on a standard laptop. Gemini 3.1 Pro (February) produced notably superior visual outputs. The gap between frontier and open models is compressing faster than expected, with practical implications for local AI development and privacy-sensitive deployments.