LLM · Google

Gemma 4 26B

Google's 2 April 2026 Gemma 4 post describes the 26B model as a mixture of experts that activates 3.8 billion parameters. It says the larger models offer up to a 256K context window, the weights are Apache 2.0, and they can be downloaded from Hugging Face, Kaggle, or Ollama. The post does not publish a token price.

Checked 19 Oct 2026https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/