Development
An encoder-free multimodal model runs on a laptop under Apache 2.0
Claim
Google's Gemma 4 12B pipes vision and audio straight into the LLM backbone, runs on a laptop and nearly matches its 26B Mixture-of-Experts sibling at under half the memory, all under Apache 2.0, as the family passed 150 million downloads.
Reported metric: 150M downloads
Reporting date: (not necessarily the event date)
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to June 4, 2026 (theinnermostloop.substack.com)
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.