Building super fast experiences with Gemma just got easier. Gemma 4 MTP is now officially merged into llama.cpp. Developers can now pair MTP with Gemma 4 QAT for a fast, lightweight setup. https://twitter.com/googlegemma/status/2064030477628182814/photo/1
中文: 与Gemma合作打造超快体验变得更加简单。 Gemma 4 MTP 现已正式合并为 lamama.cpp。开发者现在可以将MTP与Gemma 4 QAT配对,以快速、轻便地进行设置。
We just dropped Gemma 4 Quantization-Aware Training (QAT) checkpoints on Hugging Face! All Gemma 4 model sizes and their drafters are now optimized with QAT to cut memory requirements and maximize on-device performance!
中文: 我们刚刚在拥抱面上投放了Gemma 4 Quantization-Aware训练(QAT)检查点! 所有Gemma 4型号及其绘图器现已通过QAT优化,以降低内存需求并最大化设备性能!
Introducing Magenta RealTime 2, a new open model musicians can play as an instrument! Run low-latency, live music synthesis natively on your MacBook using MIDI, text, and audio. 🎶 We love seeing Google’s open model ecosystem grow! https://twitter.com/googlegemma/status/2062619217967628693/video/1
中文: 推出Magenta RealTime 2,一款新的开放式音乐模特,可作为乐器演奏! 使用 MIDI、文字和音频,在 MacBook 上运行低延迟的实时音乐合成功能。🎶 我们很高兴看到谷歌开放的模型生态系统不断发展壮大!
Meet Gemma 4 12B! A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license. Bridging the gap between edge efficiency and advanced reasoning. Here is what’s new with Gemma 4 12B: 👇 https://twitter.com/googlegemma/status/2062202706882883696/photo/1
中文: 认识一下Gemma 4 12B! 一款统一、无编码器的多模式模型,旨在将高性能智能直接带入您的笔记本电脑,并基于 Apache 2.0 许可证发布。 弥合边缘效率与高级推理之间的差距。以下是Gemma 4 12B的新功能:👇
AI video editors can't edit what isn't indexed. Learn how a developer used Gemma 4 31B locally on a 5-year-old laptop to process and index a year of raw, unlabeled video, making it fully searchable. A great look at building local-first tools. https://twitter.com/googlegemma/status/2061847854025703827/photo/1
中文: 人工智能视频编辑无法编辑未收录的内容。了解开发者如何在一台5年制笔记本电脑上本地使用Gemma 4 31B,以处理和索引一年的原始无标签视频,使其完全可搜索。打造以本地为先的工具的绝佳方式。
The first iteration of gemma-skills is officially out! 🛠️ It enables agents to build with Gemma, including using MTP to improve speed, choosing the right model size for your use case, and locating up-to-date resources. https://twitter.com/googlegemma/status/2061568650138747363/photo/1
中文: 首次推出 gemma-skills 已正式推出!🛠️ 它使代理能够使用Gemma进行构建,包括使用MTP提高速度、为使用场地选择合适的型号尺寸,以及查找最新资源。
A completely local agent that lives right inside your pocket. 📱 Watch Gemma 4 run 100% locally in the Google AI Edge Gallery app. It converts images into JSON schemas, transcribes audio, and uses agent skills to interact with apps, all entirely offline. https://twitter.com/googlegemma/status/2060411370139795877/video/1
中文: 一个完全本地的代理人,直接住在你的口袋里。📱 观看 Gemma 4 在 Google AI Edge Gallery 应用中100%本地运行。它将图像转换为JSON模式,转录音频,并使用代理技能与应用程序完全离线进行交互。
Honored to see Gemma 4 31B on TERMS-Bench, a benchmark for LLM negotiation agents based on economic negotiation! 🤝 - Environment verifies outcomes (no LLM-as-judge) - Top open-weight model alongside frontier peers - Allow diagnosing why and where agents fail https://twitter.com/googlegemma/status/2060036593591959854/photo/1
中文: 很荣幸在Terstn-Bench上看到Gemma 4 31B,这是一个基于经济谈判的LLM谈判代理基准!🤝 - 环境核实结果(无法学硕士(LLM 作为评判)) - 顶级敞标车型与前沿车型并列 - 允许诊断代理故障的原因和位置:
Introducing the newest Coral board, for efficient, on-device AI! Check out the demos in the video: - On-board speech translation - Natural language controlling hardware - Vision & sound generating music https://twitter.com/googlegemma/status/2059740184930074758/video/1
中文: 推出最新的Coral设备板,实现高效、设备上的AI! 查看视频中的演示: - 机上语音翻译 - 自然语言控制硬件 - 视觉与音效;声音生成音乐
Gemma 4 meets retro gaming! 🕹️✨ Introducing AIventure: an open-source dungeon crawler designed as a developer masterclass. Learn to build agentic workflows and integrate vibe-coding, like prompting an NPC to build a web app on the fly. https://twitter.com/googlegemma/status/2059343174112522532/video/1
中文: Gemma 4 与复古游戏相遇!🕹️✨ 推出AIventure:一款开源地牢爬虫,设计为开发者大师级。学习构建独立的工作流程并集成氛围编码,例如促使NPC即时构建网页应用。
Last week we went out into the wilderness, far from any cell service or Wi-Fi, to test an experimental Gemma 4 app, all 100% offline: 🔍Visual understanding 🧠Math reasoning 🛠️Tool use and more Running seamlessly on Pixel hardware and paired with our prototype display glasses. https://twitter.com/googlegemma/status/2057875854672515094/video/1
中文: 上周,我们前往远离任何蜂窝基站服务或Wi-Fi的荒野,测试一款可测试的Gemma 4应用程序,全部完全离线: 🔍 视觉理解 🧠 数学推理 🛠️ 使用和更多 在 Pixel 硬件上无缝运行,并搭配我们的原型展示眼镜。
We are entering a new era of on-device automation. ✨ Watch Gemma 4 E4B navigate and drive an iOS simulator directly using Argent. Local models can handle complex interactions and software navigation autonomously. https://twitter.com/googlegemma/status/2057570113390551452/video/1
中文: 我们正进入一个设备自动化的新时代。✨ 直接使用 Argent 观看 Gemma 4 E4B 导航并驱动 iOS 模拟器。本地模型可以自主处理复杂的交互和软件导航。
At Google I/O today? Stop by the Gemma Playground and chat with our Gemma 4-powered Open Duck robot! 🤖🦆 Not on site? No problem! You can 3D print your own at home. https://twitter.com/googlegemma/status/2057142732494352689/video/1
中文: 今天在谷歌I/O? 在Gemma游乐场停留,与我们的Gemma四人开放式鸭机器人聊天吧!🤖🦆 不在现场?没问题!你可以在家自费打印。
Gemma 4 is now available in ComfyUI 👇
中文: Gemma 4 现已在 ComfyUI 👇 中提供
Case study time! Hirundo trained Gemma 4 E4B to resist adversarial overrides while overcoming the alignment tax: - Weight-level defense based on Gemma 4 architecture - Stronger security posture than models over 100x its size - Preserves utility across benchmarks https://twitter.com/googlegemma/status/2054995526916227478/photo/1
中文: 案例研究时间!希伦多训练了Gemma 4 E4B,以抵制对抗性覆盖,同时克服对齐税。 - 基于Gemma 4架构的重量级防御 - 比超过其尺寸100倍的型号更坚固的安全姿态 - 在基准测试中保存实用工具
Case study time! Hirundo trains their Gemma 4 E4B variant to resist adversarial overrides, while overcoming the alignment tax: - Weight-level defense based on Gemma 4 architecture - Stronger security posture than models over 100x its size - Preserves utility across benchmarks https://twitter.com/googlegemma/status/2054992708654284844/photo/1
中文: 案例研究时间!希伦多训练他们的Gemma 4 E4B版本,以抵御对抗性覆盖,同时克服对位税: - 基于Gemma 4架构的重量级防御 - 比超过其尺寸100倍的型号更坚固的安全姿态 - 在基准测试中保留实用工具
The latest ChessBench results are out. ✨ It's fun to see Gemma 4 ranking in the top 3! https://twitter.com/googlegemma/status/2054283302090277123/photo/1
中文: 最新的国际象棋队结果已公布。✨ 看到Gemma 4位列前三,真是太有趣了!
Gemma 4 is a highly-capable model in Japanese! Amazing to see great results for it on the latest Swallow Leaderboard v2 results. ✨ Excited to see what’s next in the open model space! https://twitter.com/googlegemma/status/2053911525442928860/photo/1
中文: Gemma 4 是一款在日本功能极强的型号! 在最新的燕子排行榜 v2 上看到出色的效果,真是太棒了。✨ 很高兴看到开放模型空间接下来会发生什么!
Gemma 4 up to 3x faster, directly in your phone! 🚀 Check out the difference Speculative Decoding makes! Multi-Token Prediction (MTP) is supercharging inference speeds for Gemma 4. https://twitter.com/googlegemma/status/2052468624657654194/photo/1
中文: Gemma 4 速度最高可达 3 倍,直接在手机中!🚀 看看推测解码带来的区别吧!多代币预测(MTP)是Gemma 4的超强推理速度。
Gemma 4 just got even faster! We're releasing Multi-Token Prediction (MTP) drafters that deliver up to a 3x speedup, without any degradation in output quality or reasoning logic. https://twitter.com/googlegemma/status/2051713412431007808/photo/1
中文: Gemma 4 刚刚变得更快了! 我们正在发布多代币预测(MTP)绘图器,可实现高达3倍的加速,但输出质量或推理逻辑不会出现任何退化。
https://x.com/i/article/2049591001967443968
Gemma 🤝 React Native📱 Exciting news for mobile developers! We love seeing the community unlock new ways to build. You'll soon be able to run Gemma 4 completely on-device in React Native. https://twitter.com/googlegemma/status/2051350655135297564/video/1
中文: 杰玛 🤝 React Native📱 对移动端开发者来说,这是个令人兴奋的消息!我们喜欢看到社区开启新的建设方式。 很快你就能在 React Native 中完全在设备上运行 Gemma 4。
Curious what Gemma 4 can do in a local agentic setup? This notebook uses Haystack to walk through four practical demos: • A multimodal map/weather agent • Dynamic tool discovery via a GitHub MCP server • Visual Question Answering • RAG implementations https://twitter.com/googlegemma/status/2050328457989743001/photo/1
中文: 好奇Gemma 4在本地代理环境中能做些什么? 本笔记本使用 Haystack 来浏览四个实用演示: • 多模态地图/天气代理 • 通过 GitHub MCP 服务器进行动态工具发现 • 视觉问答 • RAG 实现
Now you can use Gemma directly in the Gemini CLI! 🚀 v0.40.0 introduces experimental support for local Gemma models, starting with intelligent model routing (with full local execution on the roadmap!). https://twitter.com/googlegemma/status/2049926903830921448/video/1
中文: 现在你可以直接在Gemini CLI中使用Gemma!🚀 v0.40.0 引入了对本地 Gemma 模型的实验性支持,首先从智能模型路由(在路线图上完全执行本地执行!)。
Offline vibe coding with Gemma 4. This open-source Electron app runs entirely on your Mac via Apple's MLX framework. No Wi-Fi required. https://twitter.com/googlegemma/status/2049585589306224958/video/1
中文: 使用Gemma 4进行离线氛围编程。 这款开源的 Electron 应用完全通过苹果的 MLX 框架在 Mac 上运行。 无需Wi-Fi。
Learn how to run a local coding agent! Use: - Pi agent - Gemma 4 26B - Serving engine of choice: e.g. LM Studio https://twitter.com/googlegemma/status/2049163687639007451/photo/1
中文: 学习如何运营本地编码代理!使用: - 管子代理 - 杰玛 4 垒打 - 服务首选引擎:例如LM Studio
A completely local agent that lives right inside your browser. Powered by Gemma 4 E2B and WebGPU, it uses native tool calling to: 🔍 Search browsing history 📄 Read and summarize pages 🔗 Manage tabs 100% local. No servers needed! https://twitter.com/googlegemma/status/2048805789788413984/video/1
中文: 一个完全本地的代理,直接住在你的浏览器里。 由 Gemma 4 E2B 和 WebGPU 提供支持,它使用原生工具调用: 🔍 搜索浏览历史 📄 阅读并总结页面 🔗 管理标签页 100%本地。无需服务器!
https://x.com/i/article/2044596881465233408
Unplugging completely! No WiFi and zero notifications. A great way to get deep focus on a project. Here is a walkthrough showing how to run Gemma 4 (26B A4B) fully offline with LM Studio & OpenCode to parse PDFs, ask questions, and build sites 100% locally. https://twitter.com/googlegemma/status/2047406475631624285/video/1
中文: 彻底拔掉插头!无WiFi和零通知。深入关注项目的好方法。 以下是如何使用LM Studio &OpenCode在线方式完全离线运行Gemma 4(26B A4B)的演练,可在本地100%解析PDF文件、提问和构建网站。
We had a fantastic time connecting with the developer community at AI Engineer Europe, in London! The team showcased our Gemma and Gemini models in action. We also shared fun, hands-on demos like the Duck Gemma prototype and the Nano Banana Bot 3000 sticker machine. https://twitter.com/googlegemma/status/2047021491796201670/video/1
中文: 我们与位于伦敦的AI Engineer Europe的开发者社区取得了一段非常愉快的时光!团队展示了我们的Gemma和Gemini模型。我们还分享了有趣的动手演示,比如Duck Gemma原型机和Nano Banana Bot 3000贴纸机。
What does it take to run 3, 5, or even 10 concurrent instances of Gemma 4 locally? We've open-sourced a demo letting you run multiple models side-by-side on your hardware. Gemma 4 26B A4B easily runs 10+ concurrent requests on a MacBook Pro M4 Max at 18 tokens/sec per request. https://twitter.com/googlegemma/status/2046621841146671456/video/1
中文: 本地运行3、5甚至10个并发的Gemma 4实例需要什么? 我们开源了一个演示,可让您在硬件上并排运行多个模型。 Gemma 4 26B A4B 可在 MacBook Pro M4 Max 上按 18 个令牌/秒要求轻松运行 10 个多个并发请求。
We are honored to be featured in the latest Two Minute Papers video! Here’s a short clip from it. https://twitter.com/googlegemma/status/2046302122325795295/video/1
中文: 我们很荣幸能出现在最新的《两分钟论文》视频中! 以下是其中的一段短片。
Check out this video on how to run Gemma 4 locally on an iPhone! It runs completely offline and handles long context, meaning no data plan, no API calls, and no monthly fees required. https://twitter.com/googlegemma/status/2045204738720084191/video/1
中文: 观看这段关于如何在iPhone上本地运行Gemma 4的视频吧! 它完全离线运行,处理长上下文,意味着无需数据计划、无需API调用,也无需月费。
Calling all AI builders in SF! 🌉 Join Google DeepMind's Gemma team and friends for a Demo Night celebrating Gemma 4! Hear from: ✨Gemma team ✨Awni Hannun (MLX) ✨Daniel Han (Unsloth) ✨Sam Herring (Nous Research) ✨Roger Wang (vLLM) ✨Henry Ndubuaku (Cactus) https://twitter.com/googlegemma/status/2044811740693864578/photo/1
中文: 给旧金山的所有人工智能构建者打电话!🌉 与谷歌DeepMind的Gemma团队及朋友们共同参加一个庆祝Gemma 4的演示之夜! 从以下内容中听到: ✨ 杰米马团队 ✨ 阿维尼·汉努(MLX) ✨ 丹尼尔·韩(解说) ✨ 萨姆·赫林(努斯研究) ✨ 王罗杰(vLLM) ✨ 亨利·恩杜布亚库(仙人掌)
Check out this amazing combo using Gemma4 + Falcon Perception for video tracking! 1️⃣Give Gemma 4 video frames 2️⃣It describes what it sees 3️⃣Falcon Perception takes those descriptions, segments the objects, and tracks them across the video! The best part? All running locally! https://twitter.com/googlegemma/status/2044535184494891250/video/1
中文: 使用 Gemma4 + Falcon Perceptions 观看视频追踪,观看这款精彩组合吧! 1️⃣ 给Gemma 4个视频帧 2️⃣ 描述它所看到的 3️⃣ 猎鹰感知这些描述,分割物体,并在视频中追踪它们! 最棒的部分?一切奔向本地!
We are in the era of local AI orchestration Gemma 4 evaluates a scene, reasons about what to ask, and calls a segmentation model to execute the vision tasks: 🚗 "Segment all vehicles." ➔ 64 found 🚙 "Now just the white ones." ➔ 23 found All happening offline on a laptop. https://twitter.com/googlegemma/status/2044138565576339545/video/1
中文: 我们正处于本地人工智能编排的时代 Gemma 4 评估一个场景,评估该问什么的原因,并调用细分模型来执行视觉任务: 🚗 “所有车辆的分类。” ➔ 64 🚙 “现在只有白人。” ➔ 23 发现 所有内容均在笔记本电脑上离线进行。
We hope you’re having fun with Gemma 4! Put its capabilities to the test and build AI that helps people in the Gemma 4 Good Challenge. https://twitter.com/googlegemma/status/2043011779983515683/photo/1
中文: 希望您能与Gemma 4一起玩得开心! 将其能力付诸验证,并构建有助于人们参加Gemma 4“好挑战”的人工智能。
We’re taking Gemma 4 on a world tour with 50+ events across the globe! From hackathons to deep dives, come build with the latest open weights in a city near you 🌎 Where we’ll be next: San Francisco, São Paulo, Kolkata, Las Vegas, Seoul, online, and more! https://twitter.com/googlegemma/status/2042682920897515636/photo/1
中文: 我们将带Gemma 4参加全球50多场活动的世界巡演!从黑客马拉松到深潜,在您附近的城市中,以最新的开放力量进行打造 🌎 接下来我们将前往:旧金山、圣保罗、加尔各答、拉斯维加斯、首尔,以及线上活动!
Gemma 4 has already been downloaded over 10 million times, with over 1,000 variants published by the community! What is your favorite Gemma 4 variant? 👇 https://twitter.com/googlegemma/status/2042310042343653509/photo/1
中文: Gemma 4 已被下载超过 1000 万次,社区发布了 1000 多种版本! 你最喜欢的Gemma 4版本是什么?👇
We love seeing what you build with Gemma! 🚀 Like this incredible community project that helps you fine-tune Gemma 4 with audio, text, and images on Apple silicon. 🧵👇 https://twitter.com/googlegemma/status/2041942389745709213/photo/1
中文: 我们喜欢看你用Gemma打造的东西!🚀 就像这个出色的社区项目,可帮助你在苹果硅片上用音频、文字和图像微调Gemma 4。🧵👇
https://x.com/i/article/2041504032410222593
Gemma 4 can run on phones without an internet connection! 🤯 It can perform local agentic tasks, such as logging and analyzing trends. When connected, it can also make API calls. Want to try it yourself? Get the Google AI Edge App on iOS or Android. (🔊 Sound on for the demo!) https://twitter.com/googlegemma/status/2041256042882105666/video/1
中文: Gemma 4 可以在没有网络连接的情况下使用手机运行!🤯 它可以执行本地代理任务,例如记录和分析趋势。连接时,它也可以进行API调用。 想自己试试吗?在 iOS 或安卓设备上获取谷歌AI Edge应用程序。(🔊 为演示提供声音!
Gemma 4 was released 3 days ago. What have you built with it so far? Show us what you got! 👇
中文: 杰玛4号于3天前发布。到目前为止,你用它建了什么? 告诉我们你得到了什么!👇
Who wants to know how Gemma 4 works? This visual guide breaks down the new architectures and how they process text, images, and (for the smaller models) audio. 👇 https://twitter.com/googlegemma/status/2040449610297323681/photo/1
中文: 谁想知道Gemma 4是如何运作的? 本视觉指南将新架构及其处理文本、图像以及(用于较小模型)的音频方式进行细分。 👇
Thanks for following us! We're excited to see what you all build with Gemma 4! In case you missed it, you can find all our checkpoints, with an Apache 2.0 License, on Hugging Face: https://twitter.com/googlegemma/status/2040107948010242075/video/1
中文: 感谢您关注我们! 我们很期待看到大家用Gemma 4打造什么! 如果漏了,您可以在“拥抱面”上找到我们所有带有 Apache 2.0 许可证的检查点:
Meet Gemma 4! Purpose-built for advanced reasoning and agentic workflows on the hardware you own, and released under an Apache 2.0 license. We listened to invaluable community feedback in developing these models. Here is what makes Gemma 4 our most capable open models yet: 👇 https://twitter.com/googlegemma/status/2039736504822763534/photo/1
中文: 认识Gemma 4! 专为您拥有的硬件上的高级推理和代理工作流而设计,并基于 Apache 2.0 许可证发布。 我们在开发这些模型时听取了社区的宝贵反馈。让Gemma 4成为我们迄今为止最具开放功能的型号:👇