Google has launched Gemma 4

"The four variants are the Effective 2B (E2B) and Effective 4B (E4B) edge models, designed to run on-device on phones, Raspberry Pi, and Jetson Nano hardware developed in collaboration with the Pixel team, Qualcomm, and MediaTek."

"Google claims both larger models outcompete models up to 20 times their size on that benchmark, with the 31B Dense model currently ranking third among all open models on the Arena AI text leaderboard."

"All four models are multimodal, natively processing video and images, and are trained across more than 140 languages, with context windows of 128K tokens for edge models and 256K for larger variants."

Gemma 4 includes a 2B edge model for Raspberry Pi and a 31B dense model ranked third on the Arena AI leaderboard. Built on Gemini 3 technology, these models are released under a more permissive Apache 2.0 license. The four variants include Effective 2B and 4B models for on-device use, and 26B and 31B models for offline use. Both larger models outperform competitors up to 20 times their size. All models are multimodal, supporting video, images, and audio input, with extensive language training and improved reasoning capabilities.

Read at TNW | Google