Google DeepMind Launches Gemma 4 Open Models with Apache 2.0 License
Curated by Ayaan Zaveri
Apr 2, 2026
Artificial Intelligence
Google DeepMind released Gemma 4, the most capable open AI model series yet, offering advanced reasoning and on-device agentic workflows for local and edge deployments blog.google+1.
The models are fully open-source under the Apache 2.0 license, simplifying access and encouraging a vibrant ecosystem of autonomous AI experiences zdnet.com+1.
Gemma 4 supports multimodality (text, image, audio), 140+ languages, and a 256,000-token context window, enabling complex conversations, code analysis, and large-scale data processing zdnet.com+1.
The family includes four variants—2B Edge, 4B, 26B MoE, and 31B Dense—optimized for resource-constrained devices to workstation-class workloads, balancing intelligence with computational efficiency huggingface.co+1.
By enabling on-device inference, Gemma 4 delivers offline functionality, enhanced data privacy, and lower operational costs, especially valuable for regulated or cost-sensitive use cases zdnet.com+1.
Sources
7 references
- Gemma 4: Byte for byte, the most capable open models - blog.google · Apr 1
- Bring state-of-the-art agentic skills to the edge with Gemma 4 - Google Developers Blog · Apr 1
- Google's Gemma 4 model goes fully open-source and unlocks powerful local AI - even on phones - zdnet.com · Apr 2