> gemma 4

Google's open-weights model family, Apache 2.0 — run it with Ollama or llama.cpp, serve it as an OpenAI-compatible API, steer thinking mode, and fine-tune on your own data