Gemma is Google DeepMind's family of open-weight models, positioned as Google's open-weight offering alongside Meta's Llama and Mistral's open-weight releases. It is distinct from Google's closed-source Gemini family, with a different architecture, different weights, and a different license: Gemma weights are released openly, while Gemini is closed. Google has described Gemma as powering enterprise fine-tuning and on-device deployments.
Generations
The family has advanced through several generations since its first release:
- Gemma 1 (February 2024): 2B and 7B variants.
- Gemma 2 (2024): 9B and 27B variants.
- Gemma 3 (2025): additional variants and multimodal extensions.
- Gemma 4 12B (around June 2, 2026): an encoder-free, unified multimodal open model released under an Apache 2.0 license. It fits in 16GB of memory and runs agentic reasoning and native audio on a laptop. According to the cited report, it approaches the performance of Google's larger 26B mixture-of-experts model at less than half the memory footprint. It was released in the same early-June window as NVIDIA's open Nemotron 3 Ultra (550B) and Microsoft's MAI slate. (Source: nlp.elvissaravia.com)
- DiffusionGemma (June 10, 2026): an experimental open 26B mixture-of-experts model (3.8B active parameters) built on a Gemma 4 backbone that uses text diffusion to generate blocks of text in parallel rather than strictly left-to-right, reported to deliver up to roughly 4× faster GPU inference (Source: blog.google; siliconangle.com).
A weekly agents-research roundup for the week ending July 11, 2026 reported that Google open-sourced Gemma 4 that week, listing it alongside Tencent's open-weight Hy3 release (Source: nlp.elvissaravia.com); the relationship of this release to the June Gemma 4 12B release is not specified in the available reporting.
Evaluations
Gemma 1.1 2b IT, the instruction-tuned variant, was rated the least politically biased conversational LLM in an analysis by David Rozado published in January 2025 (Manhattan Institute — Measuring Political Preferences in AI Systems (Rozado)).
Relationships
- developer: Google DeepMind
- sibling-family: Gemini 3 / Gemini 3 Pro, Gemini 2.5 Pro (closed-source Gemini family)
- related: Open-Source AI / Open-Weight Models, Llama 4 (Meta), Manhattan Institute — Measuring Political Preferences in AI Systems (Rozado).