Get the latest crypto news, updates, and reports by subscribing to our free newsletter.
Giấy phép số 4978/GP-TTĐT do Sở Thông tin và Truyền thông Hà Nội cấp ngày 14 tháng 10 năm 2019 / Giấy phép SĐ, BS GP ICP số 2107/GP-TTĐT do Sở TTTT Hà Nội cấp ngày 13/7/2022.
© 2026 Index.vn
Google has unveiled Gemma 4, an open-source AI model positioned to improve reasoning performance and enable smart automation systems. The company says the model is designed to support multi-step planning, deep logical reasoning, and the development of AI agents that can interact directly with tools and APIs.
Gemma 4 is currently available in four versions, intended to run across a range of hardware configurations, from Android smartphones and laptop GPUs to workstations and accelerated systems. Google highlights that the model can plan multiple steps and supports code generation.
The release also emphasizes multimodal processing, with support for handling both images and video. Google describes the model as capable of operating in environments that require more than single-step responses, including tasks that benefit from structured reasoning.
Google says the lighter Gemma 4 versions have a context window of 128,000 tokens, while the larger versions support up to 256,000 tokens. The company also notes that the model has been trained on more than 140 languages.
Google reports that since its launch, Gemma has been downloaded more than 400 million times. It also says the Gemmaverse developer community has created over 100,000 variants, pointing to continued momentum around the open AI ecosystem.

Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…