Google Gemma 4 Adopts Apache 2.0 License: A True Open Source Frontier Model

Google has switched Gemma 4 from its restrictive custom license to the Apache 2.0 license — the same permissive license used for Android and Kubernetes — making it one of the most freely available frontier AI models.

What Changed

Competitive Position

Gemma 4 now directly competes with:

• Meta Llama 3: Uses custom Llama License (more restrictive than Apache 2.0)
• Mistral: Already Apache 2.0, but fewer model sizes
• DeepSeek V3: MIT license, but much larger compute requirements

Why Apache 2.0 Matters

Developers can now:

• Use Gemma 4 in commercial products without revenue limits
• Build derivative models and services
• Deploy on edge devices (E2B/E4B run on phones and Raspberry Pi)
• Integrate into proprietary software stacks

This is especially significant for the E2B and E4B models designed for mobile and IoT, where open licensing removes barriers to embedded AI adoption.