88Z AI, 3 April — Google has released the Gemma 4 series of open models, featuring E2B, E4B, 26B MoE, and 31B Dense versions. Designed for advanced reasoning and agent workflows, this series is touted as the most powerful in its class, supporting multimodality and long contexts under an Apache 2.0 open-source license.
At a glance
- The Gemma 4 series includes four versions designed for advanced reasoning and agent workflows, ranking highly in performance for their size.
- All models support multimodality, long contexts, and over 140 languages, and are open-sourced under the Apache 2.0 license.
The update
Google has officially launched the Gemma 4 series of open models, aiming to deliver the strongest performance within their respective size classes. The series includes four specific versions: E2B, E4B, 26B Mixture-of-Experts (MoE), and 31B Dense, all optimized for advanced reasoning tasks and agent workflows. In terms of performance evaluation, the 31B and 26B models have achieved notable results on the Arena AI leaderboard, ranking third and sixth globally among open models, respectively. The entire Gemma 4 lineup features multimodal capabilities, supports long context windows, and covers over 140 languages. Furthermore, these models are open-sourced under the Apache 2.0 license, providing developers with flexible usage rights.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 谷歌开发者Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.