"Google's Gemma 4 family includes four versions, with the 31-billion and 26-billion models achieving high rankings on Arena AI's text leaderboard, outperforming models 20 times their size."
"The Gemma 4 models can process video and images, making them suitable for tasks like optical character recognition, and the smaller models can also handle audio inputs."
"With the release of Gemma 4 under the Apache 2.0 license, Google aims to provide greater accessibility for developers, moving away from its previous proprietary licensing."
"Google claims that the Gemma 4 family has achieved an unprecedented level of intelligence-per-parameter, indicating significant advancements in the efficiency of large language models."
Google has introduced the Gemma 4 family of open-weight models, available under the Apache 2.0 license. This release includes four versions, varying by parameters: 2-billion and 4-billion models for edge devices, and 26-billion and 31-billion models for powerful machines. The models excel in processing video, images, and audio, and can generate offline code. Google claims these models demonstrate an unprecedented intelligence-per-parameter ratio, outperforming larger models on AI leaderboards. They support over 140 languages, enhancing their utility for diverse applications.
Read at Engadget
Unable to calculate read time
Collection
[
|
...
]