Google introduced Gemma 4 with stronger local and edge performance. But the bigger move is licensing: Gemma now ships under Apache 2.0 instead of a restrictive custom license.
That change removes one of the biggest blockers that kept many teams from deploying Gemma in production. For enterprises, legal clarity often matters as much as model quality—and sometimes more.
Gemma 4 spans small edge variants and larger workstation-class models, with Google positioning it for reasoning-heavy and agentic workflows. In practice, this feels less like “here’s a faster model” and more like “here’s a model stack you can actually ship.”
Why this matters
- Lower legal friction: Apache 2.0 is familiar, enterprise-friendly, and procurement-safe.
- Better local-first AI options: stronger offline/on-device paths for coding, OCR, and assistant workflows.
- Competitive pressure: this raises the bar for “open” model providers still using restrictive terms.
The strategic implication is clear: open models are entering a new phase where licensing, not just leaderboard performance, decides adoption. Google is trying to win that layer.