Google Opens Gemma 4 Under Apache 2.0 with Multimodal and Agentic Capabilities

Google has released Gemma 4, a family of open-weight AI models (2B, 4B, 26B MoE, and 31B dense) under the Apache 2.0 license. Key capabilities include native video, image, and audio processing, context windows up to 256K tokens, function-calling for agentic workflows, and multilingual support across 140+ languages. The 31B dense model scores 84.3% on GPQA Diamond and 80.0% on LiveCodeBench v6, outperforming models three to five times its size. The Apache 2.0 license allows unrestricted commercial use, fine-tuning, and modification. Weights are available day-zero on Hugging Face, Kaggle, Ollama, llama.cpp, LM Studio, and other platforms.