
Ep. 187 | Google Just Shipped an AI That Runs Entirely on Your Laptop
Google released Gemma 4 12B, a unified multimodal model that runs locally on consumer laptops with 16 gigabytes of memory or VRAM. Unlike previous multimodal models that required cloud infrastructure and separate vision encoders, Gemma 4 12B is encoder-free — it receives text, images, video, and native audio inputs directly into a single model backbone without preprocessing pipelines. It runs under Apache 2.0 license and is part of a broader on-device toolkit including Google AI Edge Gallery and LiteRT-LM serve.
Michael and Frank break down why this is a genuine inflection point for small…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.