#google

Google Tests Local AI Demand With Gemma 4 12B Release

Google released Gemma 4 12B as an open-weights multimodal AI model designed to run locally on a standard enterprise laptop. The model is described as an 11.95-billion-parameter system with an Apache 2.0 license, 16GB memory target, 256K context window and immediate availability through Google AI Edge Gallery. The practical test is whether enterprises use local multimodal inference when cloud access, latency or data handling are constraints.