TL;DR: Google's Gemma 4 12B model is now supported on Ollama, enabling local execution of a native multimodal model.
Summary: Ollama has added support for Google's Gemma 4 12B model across all supported platforms. The Gemma 4 12B architecture natively integrates text, image, and audio capabilities directly without relying on separate vision or audio encoder models. Developers can run the model locally and launch it with agentic tools like Claude Code and Hermes Agent.
Why it matters: This makes it straightforward to run a highly capable, natively multimodal 12B model locally for private agentic workflows. Developers should test Gemma 4 12B locally to evaluate its latency and agentic capabilities compared to larger cloud models.
Source: @ollama