Google launches Gemma 4, a new open-source model: How to try it
agents autonomous deepmind gemma google open-source
| Source: Mashable | Original article
Google DeepMind unveiled Gemma 4 on Thursday, the latest iteration of its open‑source family of large language models. The 4‑bit‑quantised model is released under an Apache 2.0 licence and is immediately available on GitHub and Hugging Face, with ready‑to‑run Docker images and a simple “gemma‑run” CLI that lets developers spin up the model on a laptop, a Raspberry Pi or any edge device that supports the new LiteRT‑LM runtime.
Gemma 4 is pitched as the most capable open model to date, delivering “agentic” abilities such as multi‑step planning, tool use and autonomous decision‑making while supporting more than 140 languages. DeepMind highlights a 30 % jump in reasoning‑per‑parameter compared with Gemma 3, a result of a larger training corpus and a refined transformer architecture. By targeting on‑device inference, the model promises lower latency, reduced data‑center costs and stronger privacy guarantees for applications ranging from personal assistants on smartphones to predictive maintenance on industrial IoT sensors.
The release matters because it narrows the gap between proprietary offerings such as Google’s Gemini or OpenAI’s GPT‑4 and community‑driven models like Meta’s LLaMA 2 or Mistral‑7B. Developers can now experiment with high‑level reasoning without paying for cloud APIs, potentially accelerating innovation in Nordic startups that specialise in embedded AI for health tech, logistics and renewable‑energy monitoring. At the same time, the open‑source nature raises the usual safety questions about misuse and the need for robust alignment tools.
Watch for early benchmark results that will reveal how Gemma 4 stacks up against rival open models, for integration hints in upcoming Android 15 and Chrome OS releases, and for DeepMind’s promised fine‑tuning toolkit that could let enterprises customise the model for domain‑specific tasks while retaining the open‑source ethos.
Sources
Back to AIPULSEN