# Gemma: what I've logged

> Gemma comes up in 4 posts of my AI log, from September 2025 to June 2026, and each item below links to the primary source. The sources come from Google and Google DeepMind, filed under open weights, robotics and frontier.

## Items

1. **Gemma 4 12B** (Google, 1 Jun 2026): A June 3 mid-size Gemma that takes image and audio without separate encoders and runs on a 16GB laptop, under Apache 2.0. [Primary source](https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemma-4-12B/) · [Post](https://heyaliasif.com/ai/2026-06-01-week-in-ai#item-3)
2. **Gemma 4** (Google DeepMind, 30 Mar 2026): Four sizes (E2B, E4B, a 26B mixture-of-experts and a 31B dense) under Apache 2.0, with the 31B ranked third among open models on the Arena text leaderboard, per Google. [Primary source](https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/) · [Post](https://heyaliasif.com/ai/2026-03-30-week-in-ai#item-1)
3. **TranslateGemma** (Google, 12 Jan 2026): Open translation models built on Gemma, in sizes that fit on ordinary hardware. [Primary source](https://huggingface.co/google/translategemma-4b-it) · [Post](https://heyaliasif.com/ai/2026-01-12-week-in-ai#item-2)
4. **MedGemma 1.5** (Google, 12 Jan 2026): An updated open medical model from Google, small enough to run inside a hospital's own environment. [Primary source](https://huggingface.co/google/medgemma-1.5-4b-it) · [Post](https://heyaliasif.com/ai/2026-01-12-week-in-ai#item-3)
5. **VaultGemma 1B** (Google, 8 Sept 2025): A Gemma trained from scratch with differential privacy, so it can't memorize and repeat individual training records. [Primary source](https://huggingface.co/google/vaultgemma-1b) · [Post](https://heyaliasif.com/ai/2025-09-08-week-in-ai#item-4)

## Posts

- 2026-06-01 [Nemotron 3 Ultra ships, MiniMax M3, Gemma 4 12B, Glasswing grows](https://heyaliasif.com/ai/2026-06-01-week-in-ai)
- 2026-03-30 [Gemma 4 goes Apache 2.0, Qwen3.6-Plus, GLM-5V-Turbo](https://heyaliasif.com/ai/2026-03-30-week-in-ai)
- 2026-01-12 [Anthropic's Cowork, DeepSeek's Engram and a new open image editor](https://heyaliasif.com/ai/2026-01-12-week-in-ai)
- 2025-09-08 [Claude makes files, Qwen3-Next goes sparse, ERNIE 4.5 Thinking](https://heyaliasif.com/ai/2025-09-08-week-in-ai)

Canonical: https://heyaliasif.com/ai/topics/gemma