Google's latest on-device AI model is custom-made for your laptop

Google launched Gemma 4 12B, a mid-sized on-device AI model for laptops with at least 16GB RAM. It supports native audio input and uses an encoder-free architecture to deliver multimodal performance without latency. The model matches the Gemma 4 26B MoE model in benchmarks. It enables vision and audio processing through single matrix multiplication and direct signal projection into text token space.