Google launched the Gemma 4 family of open source AI models on April 2, 2026, its most ambitious open contribution yet, under the Apache 2.0 license, covering variants from 2 to 31 billion parameters. By late April 2026, Gemma had surpassed 400 million downloads and spawned over 100,000 community-created variants. Google, Meta, Microsoft, and Nvidia are united in defending open-weights models, arguing that restrictions would benefit closed-model vendors without improving safety outcomes. Anthropic and OpenAI are pushing for stricter regulatory frameworks, particularly regarding Chinese open source AI initiatives. By giving away powerful models for free, Google strengthens developers’ reliance on its cloud infrastructure and TPU hardware.
Source: Read the original article
