gemma-4-26b-a4b

Gemma 4 26B-A4B is an open-source large language model released by Google, belonging to the Gemma 4 series. It uses a Mixture of Experts (MoE) architecture with approximately 26 billion total parameters and around 4 billion active parameters. The model supports text understanding and generation tasks, can be deployed in local environments, and is applicable to a range of natural language processing scenarios including question answering, summarization, and code assistance.