GOOGLE GEMMA 4: APACHE 2.0 LICENSE CHANGES THE OPEN-SOURCE AI GAME

Google DeepMind’s release of Gemma 4 under Apache 2.0 license represents a significant shift in the open-weights AI landscape. Unlike previous “open” models with restrictive licenses, Gemma 4 offers true freedom: commercial use, modification, and distribution without royalties or attribution requirements.

Technical highlights:

• Four model sizes for different deployment needs
• Built from Gemini 3 technology (same research lineage)
• Apache 2.0 license (unrestricted commercial use)
• Competitive benchmarks with frontier models

The licensing distinction matters:

• Many “open” models restrict commercial use
• Apache 2.0 removes legal barriers to deployment
• Companies can run Gemma 4 on-premise without API costs
• Researchers can study and modify weights freely

For developers, Gemma 4 changes the build-or-buy calculus:

• Startups can ship products without per-token costs
• Enterprises can run models on internal infrastructure
• Researchers have full access to model internals
• The open-weights ecosystem is maturing rapidly

The timing — same day as Microsoft’s model announcements — suggests Google is positioning open models as an alternative to both Microsoft’s proprietary stack and OpenAI’s paid APIs. Developers who want capability without dependency now have a credible path.

Sources:

• Google DeepMind: Gemma 4 announcement
• Gigazine: Apache 2.0 license analysis
• Developer community feedback