Google has introduced Gemma 4, a set of four open-weight models licensed under Apache 2.0. The license update from earlier Gemma versions broadens the available options for modification, reuse, and commercial use.
These models are designed to support a variety of hardware, from mobile devices to server-scale inference. The release was announced by Google DeepMind and focuses on reasoning, agent-based workflows, and multimodal input support.
Gemma 4 Model Sizes and Hardware Targets
Google has announced the release of Gemma 4 in several configurations. These include E2B, which is optimized for edge devices and targets mobile hardware and low-power situations; E4B, also edge-optimized and aimed at similar devices but with higher capacity
- 26 billion parameter mixture of experts model designed for server-scale use, which currently ranks sixth on the Arena AI text leaderboard among open models
- 31 billion parameter dense model that is ranked third on the same leaderboard
Google notes that both the 31B and 26B models outperform other models with fewer parameters by up to 20 times on the Arena AI benchmark. These performance claims are based on Google’s own assessments at the time of the models’ release.
Capabilities Across the Gemma 4 Model Family
All four Gemma 4 models come with native multimodal support for video and images at various resolutions, with OCR and chart understanding listed as key use cases. The E2B and E4B models also include native audio input for speech recognition.
The context window lengths are 128K tokens for the edge models and up to 256K tokens for the 26B and 31B models. All models support function calling, structured JSON output, and native system instructions for building agent workflows. Google notes that the models were trained on over 140 languages natively.
From Custom Gemma License to Apache 2.0
Previous Gemma model releases used a custom license that restricted modification and redistribution. Gemma 4 now uses the Apache 2.0 license, offering standardized terms for commercial use, modification, and distribution without the need for a separate agreement with Google. Google states that this change makes Gemma 4 more suitable for enterprise and developer applications.
All four Gemma 4 models are now available. However, Google has not provided a timeline for when the edge models might be integrated into Android or consumer hardware products.
