Gemma 4 Has Landed!

Google has released Gemma 4, a family of four open-weight models under an Apache 2.0 license — a significant shift from previous restrictive licensing. The lineup includes two workstation models (31B dense and 26B MoE with ~4B active parameters) and two edge models (E2B and E4B). All models feature native multimodality with vision, audio, function calling, and chain-of-thought reasoning built into the architecture. The edge models support on-device ASR, speech-to-translated-text, and have a 128K context window, while workstation models offer 256K context. The audio encoder has been compressed by 50% compared to Gemma 3N. Models are available on Hugging Face and Google Cloud, with serverless deployment via Cloud Run using Nvidia RTX Pro 6000 GPUs. The Apache 2.0 license means commercial use, fine-tuning, and redistribution are now unrestricted.