Google Cloud Launches Gemma 4, Its Most Powerful Open AI Models
Google Cloud introduces Gemma 4, a highly capable family of open models built on Gemini 3 research. The models feature massive context windows, multimodal processing, and enterprise-grade deployment options.
Google Cloud releases Gemma 4 as its most capable family of open models to date. Built from the same research as Gemini 3 and available under a commercially permissive Apache 2.0 license, these models go beyond simple chat interactions. They feature context windows up to 256K, native vision and audio processing, and fluency in over 140 languages to handle complex logic, offline code generation, and agentic workflows.
These new models provide enterprise AI that executes complex logic while keeping data securely within designated boundaries. Organizations deploy Gemma 4 across Google Cloud to meet strict compliance guarantees, including Sovereign Cloud solutions. This approach establishes a strong foundation for digital sovereignty by granting teams complete control over their data, infrastructure, and models.
Developers access Gemma 4 through Vertex AI by selecting the model from Model Garden and provisioning their own compute resources. Users fine-tune the models using Vertex AI Training Clusters with optimized recipes, and a fully managed serverless version of the Gemma 4 26B MoE model arrives in Model Garden soon. Additionally, builders leverage the open-source Agent Development Kit to utilize the advanced reasoning, function calling, and code generation capabilities of these models.