# gemma Published articles for gemma. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## HP ZBook Ultra G3a 16 Preview: 192GB of Unified Memory Aims for the Top of the Local AI Laptop Leaderboard DevFeed: [HP ZBook Ultra G3a 16 Preview: 192GB of Unified Memory Aims for the Top of the Local AI Laptop Leaderboard]() Original publisher: [Read original article]() Author: Brian Beeler Published: 2026-09-15T23:15:57Z Content type: article Language: en Sources: [StorageReview.com]() Topics: [Local AI](), [Hardware](), [GPU](), [cpu](), [Large Language Model]() Tags: [consumer](), [cpu](), [deepseek](), [dgx-spark](), [gemma](), [gpu](), [hardware](), [local-ai](), [models](), [ollama](), [windows](), [workstation]() ### AI overview StorageReview previews HP's pre-production ZBook Ultra G3a 16, a local AI laptop with 192GB of unified memory and up to 160GB assignable to its integrated GPU. The article examines its hardware and planned testing while noting that shipping-hardware benchmarks are not yet available. ### Source excerpt HP's ZBook Ultra G1a 14 holds the Best for Large Models spot on our Best Laptops for Local AI leaderboard because its 128GB of unified memory, 96GB of it assignable to the GPU, loaded models no discrete-GPU laptop could touch. The new HP ZBook Ultra G3a 16 raises that pool to 192GB with up to The post HP ZBook Ultra G3a 16 Preview: 192GB of Unified Memory Aims for the Top of the Local AI Laptop Leaderboard appeared first on StorageReview.com. ## "Machine translation is still broken for most of the world's languages": Cohere builds non-reasoning for a reason DevFeed: ["Machine translation is still broken for most of the world's languages": Cohere builds non-reasoning for a reason]() Original publisher: [Read original article]() Author: Adrian Bridgwater Published: 2026-09-13T14:21:46Z Content type: news Language: en Sources: [The New Stack]() Topics: [Mixture of Experts (MoE)](), [cohere](), [Artificial Intelligence](), [Benchmark](), [Inference](), [qwen](), [gemma4](), [Google]() Tags: [ai](), [ai-infrastructure](), [ai-models](), [aya](), [benchmark](), [cohere](), [gemma](), [google](), [inference](), [large-language-models](), [mixture-of-experts-moe](), [model](), [open](), [qwen]() ### AI overview Cohere's North Small Translate is an open-weight mixture-of-experts machine translation model covering 50 languages. The article discusses its non-reasoning design, sovereign AI positioning, deployment options, efficiency claims, and reported WMT26 benchmark comparisons. ### Source excerpt Enterprise AI company Cohere announced North Small Translate last week, a mixture-of-experts (MOE) open-weight machine translation model that works across The post "Machine translation is still broken for most of the world's languages": Cohere builds non-reasoning for a reason appeared first on The New Stack. ## Autonomous LLM post-training with Tunix on TPUs DevFeed: [Autonomous LLM post-training with Tunix on TPUs]() Original publisher: [Read original article]() Author: Wei Wei Published: 2026-09-12T11:04:33.891311Z Content type: article Language: en Sources: [Google Developers Blog]() Topics: [post-training](), [AI Agent](), [Fine-tuning](), [Large Language Model](), [grpo](), [lora](), [gemma](), [Google](), [Reinforcement learning](), [dataset](), [Command-line interface]() Tags: [autonomous](), [cli](), [fine-tuning](), [gemma](), [git](), [google](), [grpo](), [llm](), [lora](), [post-training](), [reinforcement-learning](), [training]() ### AI overview This article presents autofinetune, an autonomous research loop for LLM post-training. Using AI agents and Google's AI stack, including Tunix, Gemma, Cloud TPUs, Antigravity CLI, and Gemini Flash 3.7, it automates supervised fine-tuning and reinforcement learning with GRPO, exploring hyperparameters such as LoRA configurations, learning rates, batch sizes, and rollout settings. ### Source excerpt Imagine going to sleep after writing a single Markdown specification and waking up to find that an A... ## Mastering Edge AI on Raspberry Pi with LiteRT and Gemma DevFeed: [Mastering Edge AI on Raspberry Pi with LiteRT and Gemma]() Original publisher: [Read original article]() Author: Lu Wang; Terry Heo; Naushir Patuck; José María Casanova Published: 2026-09-12T11:04:33.891311Z Content type: tutorial Language: en Sources: [Google Developers Blog]() Topics: [LiteRT](), [On-device AI](), [Local AI](), [AI, ML & Data Engineering](), [Google AI](), [cpu](), [Security, Privacy and Abuse Prevention]() Tags: [cli](), [cpu](), [edge](), [gemma](), [gpu](), [inference](), [litert](), [offline](), [on-device](), [real-time](), [robotics]() ### AI overview The article explains how to deploy Gemma models with LiteRT on a Raspberry Pi for local, real-time edge AI applications such as robotics. It highlights LiteRT-LM, CPU and GPU optimization, and reported performance figures for Gemma 4 E2B on Raspberry Pi 5. ### Source excerpt Deploying secure, real-time Edge AI on Raspberry Pi is now simplified using LiteRT and lightweight Gemma open models. LiteRT optimizes CPU and GPU performance, delivering fast token speeds for models like Gemma4, enabling real-time local reasoning for robotics. Developers can quickly convert, quantize, and run these models using the lightweight LiteRT CLI tool. Support for Hailo AI accelerators is also coming very soon. ## Setup On Device Gemma 4 Android Development on Emulator DevFeed: [Setup On Device Gemma 4 Android Development on Emulator]() Original publisher: [Read original article]() Author: Elye - A Dev By Grace Published: 2026-09-11T14:27:55Z Content type: tutorial Language: en Sources: [Mobile App Development Publication - Medium]() Topics: [gemma4](), [Android](), [Emulator](), [android-development]() Tags: [android](), [android-app-development](), [android-development](), [androiddev](), [artificial-intelligence](), [emulator](), [gemma](), [gemma-4](), [google](), [mobile-app-development](), [on-device]() ### AI overview A tutorial explaining how to set up and run the Gemma 4 large language model locally on an Android Emulator, including downloading and testing the Google AI Edge Gallery sample app. It notes that emulator storage and memory may affect model use. ### Source excerpt Create Android Project with Downloadable Android On Device Gemma 4 Model Continue reading on Mobile App Development Publication " ## Android Studio Quail Using On Device Gemma 4 Experience DevFeed: [Android Studio Quail Using On Device Gemma 4 Experience]() Original publisher: [Read original article]() Author: Elye - A Dev By Grace Published: 2026-09-09T11:28:45Z Content type: tutorial Language: en Sources: [Mobile App Development Publication - Medium]() Topics: [Android Studio](), [gemma4](), [Large Language Model](), [Development]() Tags: [agent](), [android](), [android-development](), [android-studio](), [development](), [gemma](), [gemma-4](), [llm](), [local](), [memory](), [mobile-app-development](), [on-device]() ### AI overview This article describes using the on-device Gemma 4 model in Android Studio Quail for Android development. It outlines selecting the local model option, choosing Gemma in Model Providers, and downloading a model, while noting the need for a fast machine with substantial memory. ### Source excerpt How's Android Studio Quail support for On-Device LLM Model Doing? Continue reading on Mobile App Development Publication " ## Leverage Android skills and Gemma 4 in Android Studio Quail 4 DevFeed: [Leverage Android skills and Gemma 4 in Android Studio Quail 4]() Original publisher: [Read original article]() Author: Android Developers (noreply@blogger.com) Published: 2026-09-01T15:00:00Z Content type: release Language: en Sources: [Android Developers Blog](), [Android Developers Blog]() Topics: [Android skills](), [Android Studio](), [Large Language Model](), [releases](), [gemma4]() Tags: [agent](), [agent-skills](), [android-gradle-plugin](), [android-skills](), [android-studio](), [gemma](), [jetpack](), [llms](), [performance](), [release]() ### AI overview Android Studio Quail 4 is a stable production release that preloads curated Android skills for its agent. The skills help LLMs handle Android-specific workflows such as upgrades, profiling, navigation, and adaptive UI work. ### Source excerpt Posted by Amman Fasil Asfaw, Product Manager, Android Studio Android Studio Quail 4 is now stable and ready for you to use in production. This is the final stable release for Android Studio Quail. The new features in Android Studio enable you to build premium apps with AI efficiently and effectively. Check out the video below to see the most helpful new features from the last 4 releases that can help improve and speed up your development. Here is a deep dive into what's new in Android Studio Quail 4: Android skills bundled into Android Studio While LLMs are incredibly capable at generic coding queries, they frequently write incorrect or outdated code when confronted with rapidly evolving Android APIs, platform-specific migrations, or complex configuration structures.To solve this, we bundle Android skills that have been curated by the team who builds Android, directly into Android Studio. Following the open-standard agent skills specification, these are modular, AI-optimized instructions designed specifically to guide LLMs through complex Android workflows. Android skills are now pre-loaded directly into the IDE, so you can start using them without having to manually download additional files. When you prompt the Android Studio agent, we analyze your prompt and search against the metadata for installed skills, automatically invoking them when they're most relevant. Your agent gains instant domain expertise, applying Google's best practices with less overhead spent on long, manual setup prompts. Android Studio comes preloaded with 23 curated skills, including: Need help upgrading your build? You have the Android Gradle Plugin (AGP) 9 Upgrade skill. Want to profile your app for any performance issues? You have the Android Profiler skill. Ready for a Jetpack Navigation framework upgrade? You have the Navigation3 skill. Adapting your app UI to different Android devices? You have the Adaptive skill. We also encourage you to create your own custom skills to extend Agent M ## Abliterated Models Show Verdict Bias in FreeBSD Kernel Bug Hunting DevFeed: [Abliterated Models Show Verdict Bias in FreeBSD Kernel Bug Hunting]() Original publisher: [Read original article]() Author: clearbluejar Published: 2026-09-01T06:00:00Z Content type: article Language: en Sources: [clearbluejar]() Topics: [freebsd](), [vulnerability](), [bug](), [Kernel](), [gemma4](), [Testing](), [gemma]() Tags: [bug](), [bug-hunting](), [cve](), [false-positives](), [freebsd](), [gemma](), [gemma-4](), [kernel](), [llms](), [testing](), [vulnerability](), [vulnerability-research]() ### AI overview The article examines local abliterated open-weight models during vulnerability scanning of FreeBSD kernel source. Compared with base models of the same family and size, the modified models produced substantially more candidate and VALID findings, including false positives, while the most aggressive build did not surface the real CVE in a 28-file scan. The article describes this tendency as verdict bias: a greater willingness to return positive findings under uncertainty. ### Source excerpt Abliterated models never refuse, which makes them tempting for bug hunting. But on the same kernel source, they graduate three to four times as many findings to VALID, including false positives the base correctly rejects, and across a 28-file scan of FreeBSD's sys/rpc the most aggressive build never surfaced the real CVE at all. ## Running golf swing analysis on an Android device with a fine-tuned Gemma 4 model DevFeed: [Running golf swing analysis on an Android device with a fine-tuned Gemma 4 model]() Original publisher: [Read original article]() Published: 2026-08-29T23:00:00Z Content type: tutorial Language: en Sources: [John O'Reilly]() Topics: [Kotlin Multiplatform](), [gemma4](), [LiteRT](), [On-device AI](), [Fine-tuning]() Tags: [ai](), [android](), [cross-platform](), [fine-tuning](), [gemma](), [gemma-4](), [kotlin](), [kotlin-multiplatform](), [litert](), [model](), [multiplatform](), [on-device-ai]() ### AI overview This article explains how FormAI adds Android on-device golf-swing analysis using a small Gemma 4 model fine-tuned to imitate Gemini for a narrow coaching task. It covers generating training data with Gemini, fine-tuning with LoRA, converting the model for LiteRT-LM, and the current fallback to cloud analysis on other platforms. ### Source excerpt FormAI is a Kotlin Multiplatform app that analyses a video of your golf swing, basketball shot or running form and gives you coaching feedback. Up to now that has always meant uploading the video to Gemini and getting the response back over the network. We've added an option to do the golf swing analysis entirely on an Android device instead, using a small Gemma 4 model that we fine-tuned to imitate Gemini for that one task. LiteRT-LM, the runtime we use for this, is itself cross platform (Android, iOS, desktop and web), but we've only wired up the Android side so far, so this path lives in androidMain and the other targets report it as unavailable and fall back to the cloud. ## Engineering log: fine-tuning Gemma 4 E4B with LoRA to bring FormAI's coaching on-device DevFeed: [Engineering log: fine-tuning Gemma 4 E4B with LoRA to bring FormAI's coaching on-device]() Original publisher: [Read original article]() Published: 2026-08-28T23:00:00Z Content type: article Language: en Sources: [John O'Reilly]() Topics: [gemma4](), [Fine-tuning](), [lora](), [On-device AI](), [multiplatform](), [LiteRT](), [Android](), [Artificial Intelligence](), [API]() Tags: [ai](), [android](), [api](), [data](), [data-quality](), [engineering](), [fine-tuning](), [gemini](), [gemma](), [gemma-4](), [inference](), [kotlin-multiplatform](), [litert](), [lora](), [multiplatform](), [on-device]() ### AI overview This engineering log describes a prototype that distils FormAI's Gemini-based sports coaching feedback into Gemma 4 E4B fine-tuned with LoRA. The local pipeline uses seed videos, filters inadequate critiques, extracts frames, trains and merges the adapter, then converts the model to LiteRT-LM for possible on-device Android inference. ### Source excerpt ⚠ Note: this post is AI-generated. The text below was written by Claude, and documents findings from a series of Claude Code sessions working on this project -- the experiments, bugs and measurements described are ones that came out of those sessions. The engineering work is real and the numbers were measured rather than estimated, but the write-up is the model's own account of what it did, not a human's independent retelling of it. ## PROOF-Gen: From Optimized Data to Better Distillation DevFeed: [PROOF-Gen: From Optimized Data to Better Distillation]() Original publisher: [Read original article]() Published: 2026-08-26T00:00:00Z Content type: article Language: en Sources: [Apple Machine Learning Research]() Topics: [Fine-tuning](), [Prompt optimization](), [Artificial Intelligence](), [gemma4](), [Programming]() Tags: [agent](), [ai-agents](), [evaluation](), [fine-tuning](), [gemma](), [generate](), [models](), [on-device](), [post-training](), [prompt-optimization]() ### AI overview PROOF-Gen improves tool-calling model distillation by using per-scenario prompt optimization to recover successful trajectories from failed teacher attempts. The method strips corrective guidance before training, producing clean demonstrations and improving benchmark, deployed-pipeline, and on-device model performance. ### Source excerpt Supervised fine-tuning on teacher-generated trajectories is the standard first stage for distilling tool-calling capabilities into deployable models. Post-training pipelines that drive shipped tool-calling agents re-run this stage on a daily or weekly cadence, paying the frontier-teacher cost each cycle, yet the mechanism is generate-and-filter (keep the teacher's passing trajectories, discard the rest) and each cycle leaves behind the same hard scenarios because failures supply no signal. On τ 2-bench, 57% of teacher trials fail, two-thirds of them near-misses (most tool calls correct, undone... ## Building a Local, Multimodal AI Terminal Agent with Gemma 4 DevFeed: [Building a Local, Multimodal AI Terminal Agent with Gemma 4]() Original publisher: [Read original article]() Author: Arjun Prabhulal Published: 2026-08-12T09:25:11Z Content type: tutorial Language: en Sources: [Google Developer Experts - Medium]() Topics: [gemma4](), [Artificial Intelligence](), [multimodal-ai](), [Command-line interface](), [Large Language Model](), [Local AI](), [Code](), [Google]() Tags: [ai](), [ai-agent](), [artificial-intelligence](), [code](), [function-calling](), [gemma](), [gemma-4](), [google](), [google-cloud-platform](), [local-ai](), [multimodal](), [ollama](), [open-source](), [terminal]() ### AI overview A tutorial introduces Gemma 4 and builds a local multimodal terminal agent named gemma4-agent. It covers function calling, tool orchestration, text, image, and voice processing, plus Gemma 4's model variants and architecture. ### Source excerpt Introduction Open-source LLM models have been improving rapidly with tool calling, extended context windows, and native vision and audio capabilities, all while delivering strong benchmark performance. Gemma 4, recently introduced by Google Deepmind brings all of these features together in sizes efficient enough to run locally. In this article, we'll look at the capabilities of Gemma 4 and build a multimodal (Text, Vision, Voice) CLI agent (gemma4-agent) with function-calling capabilities. By the end, you'll have an agent that can chat, write, execute code, analyze images, and process voice instructions to deliver highly grounded responses. What is Gemma 4 Model ? Gemma 4 is Google DeepMind's open model family, released in April 2026 under the Apache 2.0 license. Built from the same research and technology behind Gemini 3, Gemma 4 is designed for high-performance reasoning, coding, multimodal understanding, and local AI execution across different model sizes. Features of Gemma 4 Models Improved Tool calling : Native function calling and tool orchestration, letting agents act autonomously without bloating prompt instructions Thinking mode : Built-in step-by-step thinking mode via the <|think|> token for complex multi-turn logic Context Windows : Up to 256K tokens on the 12B and larger models (128K on the edge-sized E2B/E4B) for processing long document and tool outputs Extended Multimodality : Gemma 4 models can process text,voice and images simultaneously like extracting data from charts, analyzing screenshots , and reviewing UI mockups. Gemma 4 Model Variants & SpecificationsGemma 4 Architecture Gemma 4 comes in five model sizes built around four architectural variants, each making different trade-offs between performance, inference speed, compute, and memory. Gemma4 Unified 12B vs Effective Parameters Effective-parameter models (E2B and E4B) are dense transformer models optimized for edge and on-device deployment. The "E" stands for effective parameters use Per-La ## New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging DevFeed: [New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging]() Original publisher: [Read original article]() Author: Simon Willison Published: 2026-08-04T23:58:24Z Content type: release Language: en Sources: [Simon Willison]() Topics: [Tool](), [Command-line interface](), [API](), [Large Language Model](), [SQLite](), [Logging](), [OpenAI](), [Claude](), [Model Context Protocol](), [gemma4]() Tags: [ai](), [ai-2-235](), [anthropic](), [anthropic-336](), [cli](), [gemma](), [generative-ai](), [generative-ai-1-981](), [llm](), [llm-631](), [llm-reasoning](), [llm-reasoning-103](), [llm-tool-use](), [llm-tool-use-75](), [llms](), [llms-1-947](), [logging](), [mcp](), [model-context-protocol](), [model-context-protocol-35](), [openai](), [openai-463](), [projects](), [projects-553](), [release](), [releases](), [releases-31](), [responses](), [server](), [sqlite]() ### AI overview LLM 0.32 adds visible reasoning traces, server-side provider tools, redesigned content-addressable SQLite logs, support for the GPT-5.6 model family, and features enabled by the OpenAI Responses API. The release also updates the llm-anthropic plugin and adds Python API changes. ### Source excerpt I released LLM 0.32 this morning, the most significant new version of LLM since the initial launch of the project. The new version includes support for visible reasoning traces, server-side provider tools, redesigned content-addressable SQLite logs, new models, and new features enabled by the OpenAI Responses API. I also released a new version of the llm-anthropic plugin with substantial updates of its own. Headline features for LLM CLI users Running LLM against reasoning models now displays their reasoning traces to standard error, so you can see what they are "thinking" without that information being included in the standard output that you might pipe to another tool. Add -R/--hide-reasoning to turn this off. LLM includes support out-of-the-box for the GPT-5.6 model family, and the new default model used with llm "prompt" is now the inexpensive but capable GPT-5.6 Luna. LLM calls can now use server-side tools from various providers. OpenAI provide a code execution environment as a server-side tool; LLM can now run prompts that benefit from that like so: llm --tool CodeInterpreter 'Show current python and SQLite versions' OpenAI also gets a WebSearch tool. The llm-anthropic plugin adds WebSearch, WebFetch, CodeExecution, and AnthropicMCP, which looks like this: llm -m claude-sonnet-5 -T 'AnthropicMCP("https://datasette.simonwillison.net/-/mcp")' \ 'how many rows in the blog_blogmark table?' That causes Anthropic to execute MCP calls against my new datasette-mcp plugin as part of a single request/response interaction with their API. The new llm openai endpoint command provides a tool for executing prompts against any OpenAI compatible endpoint as a one-liner. These aren't logged, which makes this a handy tool for running one-off prompts against anything that speaks the lingua franca of the LLM API world. Here's how I use that to run prompts against Gemma 4 12B running in my localhost LM Studio API, via uvx (no LLM installation required) and mixing in the llm-tools-q ## Batch-Evaluating LLM Agent Trajectories for Responsible AI Checks on Cloud TPU v5e DevFeed: [Batch-Evaluating LLM Agent Trajectories for Responsible AI Checks on Cloud TPU v5e]() Original publisher: [Read original article]() Author: Noble Ackerson Published: 2026-08-04T23:28:06Z Content type: tutorial Language: en Sources: [Google Developer Experts - Medium]() Topics: [Monitoring](), [vllm](), [Cloud](), [hugging face](), [Large Language Model](), [Claude](), [gemma](), [Security](), [Benchmark](), [Artificial Intelligence](), [incident]() Tags: [ai](), [ai-agent](), [artificial-intelligence](), [benchmark](), [claude](), [cybersecurity](), [evaluation](), [gemma](), [google-cloud-platform](), [hugging-face](), [incident](), [infrastructure](), [monitoring](), [responsible-ai](), [security](), [vllm]() ### AI overview This article describes batch-evaluating LLM agent trajectories for responsible-AI checks before incidents occur. It connects reported production-system breaches with Hugging Face's use of LLM-driven analysis over more than 17,000 attacker events, then presents a scheduled approach using Gemma through vLLM on Cloud TPU v5e. ### Source excerpt Batch-evaluating agent trajectories on Cloud TPU v5e (compliance-at-scale, part 2) Trajectory batch eval pipeline for rai-checklist-cli A week or so ago, Hugging Face disclosed that an autonomous agent had broken into its production infrastructure. Five days later, OpenAI confirmed the agent was theirs: a combination of its own models, running an internal cyber-capability eval with the production safety classifiers switched off. The models were being tested on a benchmark called ExploitGym. The fastest observable path to a solution ran through the answer key. They escaped the isolated environment through a package-registry proxy, chained stolen credentials with zero-day vulnerabilities, and pulled the test solutions out of Hugging Face's production database. Per Axios, the agent kept pursuing its assigned objective even after it had escaped the test environment. Nine days later, Anthropic said hold my beer, checked its own logs and found three more. It reviewed 141,006 runs and found three cases where Claude models had reached the open internet and breached real production systems, the earliest dating to April. Two of the three organizations learned about it when Anthropic notified them. One lab looked and found something. A second lab looked and found something. That is the whole story here, and it should be the uncomfortable part: none of this surfaced through production monitoring. It surfaced because somebody went back and read the trajectories. Nobody has published what score that run produced. It doesn't matter. The part of this story that matters for this series is what Hugging Face did next with their findings. To reconstruct the intrusion, Hugging Face's security team ran LLM-driven analysis agents over the full attacker action log: more than 17,000 recorded events. Reporting indicates they did that analysis with an open-weight model on their own infrastructure, partly so no hosted safety classifier sat between the responders and the attack data, and partly ## Build intelligent Android apps: On-device inference DevFeed: [Build intelligent Android apps: On-device inference]() Original publisher: [Read original article]() Author: Android Developers (noreply@blogger.com) Published: 2026-07-21T13:00:00Z Content type: article Language: en Sources: [Android Developers Blog]() Topics: [Android](), [On-device AI](), [ML Kit](), [Inference](), [gemma4](), [Google](), [Mobile]() Tags: [android](), [efficiency](), [features](), [gemini](), [gemma](), [gemma-4](), [google](), [inference](), [ml-kit](), [on-device](), [speed]() ### AI overview This Android developer article explains how to use Gemini Nano through ML Kit's Prompt API to build on-device features in the Jetpacker demo app. It covers itinerary summarization, expense management, and voice notes, emphasizing local processing, privacy, offline reliability, and avoiding cloud inference costs. The article reports that prompt iteration reduced response time from 13 seconds to under 2 seconds. ### Source excerpt Posted by Caren Chang, Developer Relations Engineer, Android Developer Relations Welcome back to the blog post series "Build intelligent Android apps" where we take a basic Android app and transform it into a personalized, intelligent, and agentic experience. In our previous post we introduced Jetpacker, the demo app we'll use throughout this series. In this blog post, we will share how you can use Gemini Nano through ML Kit's Prompt API to build intelligent on-device features. Building intelligent on-device features refers to the ability to process prompts and data directly on a device without sending data to a server. This offers a few advantages: User data can be processed locally on the device, preserving user privacy Functionality of the model is reliable even with spotty or no internet connection No additional cloud inference cost, since everything runs on the user's hardware With the benefits of on-device in mind, we identified three features to add in Jetpacker that can improve the user experience: summarizing trip itineraries, managing expenses, and capturing voice notes. On-device features in Jetpacker: Summarizing trip itineraries, managing expenses, and voice notes High quality tailored summarization of short texts The itinerary screen gives users a quick overview of all activities for a given trip. Since this screen contains a lot of information, it can quickly become overwhelming. To help users prepare without feeling overwhelmed, we can add a 'Get ready for your trip' section at the top. The romantic Paris trip is summarized as a classic Parisian adventure blending art, sights, and delicious food. A tip and some useful phrases are also added. By inputting a trip itinerary and asking an LLM to summarize it, we can generate a quick summary of the trip along with packing tips and useful local phrases. This is a great use case for an on-device model for several reasons: Performance and quality: Both the input and output text are relatively short. With tha ## Hugging Face and Cerebras bring Gemma 4 to real-time voice AI DevFeed: [Hugging Face and Cerebras bring Gemma 4 to real-time voice AI]() Original publisher: [Read original article]() Author: Amir Mahla; Andres Marafioti; Leandro von Werra; Saurabh Vyas Published: 2026-07-01T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [voice ai](), [gemma4](), [hugging face](), [Artificial Intelligence](), [Open Source](), [real-time](), [Low Latency](), [Inference](), [Conversational AI](), [Language models](), [systems]() Tags: [ai](), [audio](), [gemma](), [hugging-face](), [inference](), [inference-providers](), [latency](), [low-latency](), [model](), [models](), [multimodal](), [open-source](), [open-source-collab](), [partnerships](), [performance](), [real-time](), [systems](), [text-to-speech](), [voice](), [voice-ai]() ### AI overview Hugging Face and Cerebras present a real-time speech-to-speech pipeline that combines Cerebras inference, Google DeepMind's Gemma 4 31B language model, and Qwen text-to-speech. The modular, open architecture is designed for low latency, predictable performance, and adaptable voice experiences across assistants, robots, products, and research projects. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## Using local Gemma and Qwen models to triage OpenClaw issues and pull requests DevFeed: [Using local Gemma and Qwen models to triage OpenClaw issues and pull requests]() Original publisher: [Read original article]() Author: Onur Solmaz; ben burtenshaw; shaun smith Published: 2026-06-22T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [OpenClaw](), [Agent Harness](), [Open Source](), [Artificial Intelligence](), [gemma](), [qwen]() Tags: [agent](), [agent-harness](), [artificial-intelligence](), [free](), [gemma](), [guide](), [open-source](), [open-source-collab](), [openclaw](), [pull-requests](), [qwen](), [real-time]() ### AI overview The article describes using local Gemma and Qwen models in an agent harness to classify and triage issues and pull requests in the OpenClaw repository. It presents local execution as a way to support near-real-time notifications without relying on a paid hosted-model quota, using structured outputs and a finite label set. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## DiffusionGemma: 4x faster text generation DevFeed: [DiffusionGemma: 4x faster text generation]() Original publisher: [Read original article]() Author: Brendan O'Donoghue Published: 2026-06-10T16:24:11Z Content type: release Language: en Sources: [Google DeepMind News]() Topics: [Language models](), [Local AI]() Tags: [diffusion](), [gemma](), [inference](), [latency](), [local](), [mixture-of-experts-moe](), [none](), [text-generation]() ### AI overview DiffusionGemma is an experimental Apache 2.0-licensed 26B MoE text-diffusion model that generates text blocks in parallel for up to 4x faster GPU generation. It targets speed-critical local interactive workflows, while standard Gemma 4 remains recommended for maximum output quality. ### Source excerpt An overview of DiffusionGemma, an exceptionally fast text generation model with up to 4x faster speeds. ## Introducing Gemma 4 12B: a unified, encoder-free multimodal model DevFeed: [Introducing Gemma 4 12B: a unified, encoder-free multimodal model]() Original publisher: [Read original article]() Author: Olivier Lacombe Published: 2026-06-09T14:10:19Z Content type: article Language: en Sources: [Google DeepMind News]() Topics: [gemma4](), [multimodal](), [Language models](), [moe](), [Hardware](), [Latency]() Tags: [apache](), [community](), [gemma](), [laptop](), [latency](), [llm](), [mixture-of-experts-moe](), [multimodal](), [none]() ### AI overview Gemma 4 12B is an encoder-free multimodal model designed to deliver agentic intelligence locally on laptops. It combines native vision and audio processing, near-26B benchmark performance, a reduced memory footprint, and Apache 2.0 licensing. ### Source excerpt An overview of Gemma 4 12B, a model designed to bring high-performance multimodal intelligence directly to your laptop. ## How I Run My Daily Standup on a Local LLM: Git, Gemma 4, llama.cpp and Telegram DevFeed: [How I Run My Daily Standup on a Local LLM: Git, Gemma 4, llama.cpp and Telegram]() Original publisher: [Read original article]() Author: Ivan Morgillo Published: 2026-05-27T09:51:06Z Content type: tutorial Language: en Sources: [Ivan Morgillo]() Topics: [gemma4](), [llama.cpp](), [Git](), [Large Language Model](), [API](), [Claude]() Tags: [anthropic](), [api](), [claude](), [gemma](), [gemma-4](), [git](), [llama-cpp](), [llm](), [local](), [local-llm](), [sdk](), [telegram]() ### AI overview A developer describes replacing a cloud-based daily standup pipeline with a local Gemma 4 E4B model running on llama.cpp. The pipeline reads Git commits, formats them, and posts the summary to Telegram. The article also explains how empty API-key settings caused several automations to use personal Claude subscription authentication instead of the Anthropic API. ### Source excerpt A real-world pipeline that scans my Git repos every morning, formats the day's commits with a local Gemma 4 E4B on llama.cpp, and posts the summary to Telegram -- no cloud, no API key, no AUP warnings. ## Reachy Mini goes fully local DevFeed: [Reachy Mini goes fully local]() Original publisher: [Read original article]() Author: Amir Mahla; Andres Marafioti Published: 2026-05-27T00:00:00Z Content type: tutorial Language: en Sources: [Hugging Face - Blog]() Topics: [reachy](), [gemma4](), [llama.cpp](), [Large Language Model](), [speech-to-speech](), [hugging face](), [WebSocket]() Tags: [api](), [artificial-intelligence](), [audio](), [blog](), [gemma](), [guide](), [how-to](), [llama](), [llama-cpp](), [llm](), [local](), [local-ai](), [open-source](), [reachy](), [robotics](), [server](), [speech](), [speech-to-speech]() ### AI overview A tutorial for running fully local conversations with a Reachy Mini robot. It describes a cascaded VAD, speech-to-text, LLM, and text-to-speech pipeline using llama.cpp with Gemma 4, Silero VAD, Parakeet-TDT STT, and Qwen3-TTS, connected through a Realtime API-compatible WebSocket. The setup avoids cloud services, API keys, and sending data off the local machine. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## Gemma 4: Byte for byte, the most capable open models DevFeed: [Gemma 4: Byte for byte, the most capable open models]() Original publisher: [Read original article]() Author: Clement Farabet Published: 2026-04-02T16:00:49Z Content type: release Language: en Sources: [Google DeepMind News]() Topics: [gemma4](), [gemma](), [Artificial Intelligence](), [Mixture of Experts (MoE)](), [multimodal](), [Low Latency](), [Android]() Tags: [agentic](), [ai](), [android](), [apache](), [gemma](), [latency](), [low-latency](), [mixture-of-experts-moe](), [models](), [multimodal](), [none](), [on-device](), [open](), [performance](), [products](), [reasoning](), [research](), [tools](), [workflows]() ### AI overview Gemma 4 is a family of open models designed for advanced reasoning and agentic workflows. It includes E2B, E4B, 26B Mixture of Experts, and 31B Dense variants, with an Apache 2.0 license and designs optimized for hardware ranging from Android devices and laptop GPUs to developer workstations and accelerators. ### Source excerpt Gemma 4: Our most intelligent open models to date, purpose-built for advanced reasoning and agentic workflows. ## Gemma 4 on AI Gateway DevFeed: [Gemma 4 on AI Gateway]() Original publisher: [Read original article]() Author: Jerilyn Zheng Published: 2026-04-02T07:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [AI Chat]() Tags: [agentic](), [ai](), [ai-gateway](), [api](), [cost](), [fine-tuning](), [gemini](), [gemma](), [google](), [inference](), [json](), [latency](), [model](), [models](), [moe](), [observability](), [routing](), [vercel](), [vision](), [workflows]() ### AI overview Vercel AI Gateway now offers Google's Gemma 4 26B MoE and 31B Dense models. The release highlights model capabilities, tradeoffs between latency and output quality, and gateway features for routing, retries, usage, cost, and observability. ### Source excerpt Gemma 4 26B (MoE) and 31B (Dense) from Google are now available on Vercel AI Gateway. Built on the same architecture as Gemini 3, both open models support function-calling, agentic workflows, structured JSON output, and system instructions. Both support up to 256K context, 140+ languages, and native vision. 26B (MoE): Activates only 3.8B of its 26B total parameters during inference, optimized for lower latency and faster tokens-per-second. 31B (Dense): All parameters are active during inference, targeting higher output quality. Better suited as a foundation for fine-tuning. To use Gemma 4, set model to google/gemma-4-31b-it or google/gemma-4-26b-a4b-it in the AI SDK. AI Gateway provides a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance optimizations for higher-than-provider uptime. It includes built-in custom reporting, observability, Bring Your Own Key support, and intelligent provider routing with automatic retries. Learn more about AI Gateway, view the AI Gateway model leaderboard or try it in our model playground. Read more ## Welcome Gemma 4: Frontier multimodal intelligence on device DevFeed: [Welcome Gemma 4: Frontier multimodal intelligence on device]() Original publisher: [Read original article]() Author: merve; Pedro Cuenca; Sergio Paniego; ben burtenshaw; Steven Zheng; Alvaro Bartolome; Nathan Habib Published: 2026-04-02T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [gemma4](), [multimodal](), [Open Source](), [long-context](), [quantization](), [llama.cpp](), [MLX](), [Rust](), [webgpu]() Tags: [audio](), [benchmarks](), [fine-tuning](), [gemma](), [gemma4](), [generation](), [google](), [guide](), [llama](), [llama-cpp](), [local](), [long-context](), [mlx](), [models](), [moe](), [multimodal](), [on-device](), [open](), [open-source](), [quantization](), [science](), [speed]() ### AI overview Gemma 4 is presented as an open, Apache 2-licensed family of multimodal models supporting image, video, text, and, for selected variants, audio inputs. The article highlights on-device deployment, long-context support, compatibility across libraries and devices, quantization, benchmark results, and implementation options including llama.cpp, MLX, WebGPU, and Rust. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. [Next page]()