FrontierThe story, in brief
Google Opens Gemma 4 Under Apache 2.0 with Multimodal and Agentic Capabilities
Google just open-sourced Gemma 4 with multimodal chops and 256K context. Here's why that matters for your inference costs.
The KeyNews take
Why it matters
Google's Apache 2.0 release of Gemma 4 across four parameter sizes signals aggressive competition in the open-weight model space, directly challenging Meta's Llama and other open alternatives on capabilities (video, audio, extended context) while removing commercial licensing friction.
The key facts
6 to knowGemma 4 released in 2B, 4B, 26B, and 31B parameter variants
Apache 2.0 license (permissive commercial use)
Multimodal capabilities: enhanced video and image processing
Audio input support on smaller models
Context window: up to 256K tokens
Published April 16, 2026
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Google has announced the release of Gemma 4, a series of open-weight AI models, including variants with 2B, 4B, 26B, and 31B parameters, under the Apache 2.0 license. Key features include enhanced video and image processing, audio input on smaller models, and extended context windows up to 256K…