New Language Models Released

The latest version of llm-gemini (0.30) includes three new models:

• gemini-3.1-flash-lite-preview: A lightweight preview model from Google’s Gemini family.
• gemma-4-26b-a4b-it: The 26 billion parameter version of the Gemma 4 model, optimized for Italian language tasks.
• gemma-4-31b-it: The 31 billion parameter version of Gemma 4, also optimized for Italian.

The release notes highlight that these models build upon previous iterations with improved performance and capabilities. For those interested in a deeper dive, author Simon Willison has published detailed notes on the new Gemma 4 family at https://simonwillison.net/2026/Apr/2/gemma-4/.

The addition of these models expands the llm-gemini ecosystem, providing developers and researchers with more options for natural language processing tasks.