Tag

Local Ai

All articles tagged with #local ai

technology4 days ago•19 min saved

AMD's Ryzen AI Halo: A Pint-Sized Desktop for Local AI Dev

AMD unveils the Ryzen AI Halo Developer Desktop, a $3,999 compact PC designed for local AI with Windows (or Linux) out of the box, featuring a Ryzen AI Max+ 395 CPU/GPU combo for up to 126 TOPS. It includes preinstalled Ryzen AI Developer Center and playbooks to streamline model deployment, supports simple multi-box clustering over 10Gb Ethernet, and is pitched as a plug‑and‑play AI appliance for developers and on‑prem workflows—faster to get running than Linux-heavy rivals. In hands-on testing, it went from power‑on to a working local model in about 9 minutes 38 seconds and demonstrated end-to-end image generation via playbooks in about 36 minutes, though it can be noisy under load and lacks USB-A ports.

via PCMag|

#ai-desktop #amd-ryzen-ai-halo #local-ai

technology1 month ago•4 min saved

Google's Gemma 4 12B Brings On-Device AI Within 16GB RAM Reach

Google unveils Gemma 4 12B, a 12-billion-parameter model designed to run on consumer laptops with 16GB of RAM or VRAM. It achieves near-26B performance thanks to Multi-Token Prediction (MTP) and a streamlined, memory-efficient multimodal pipeline that handles text, audio, and images without a bulky encoder. The model is intended for on-device use and can be downloaded from Kaggle or Hugging Face, with integration via tools like LM Studio and Google AI Edge Gallery.

via Ars Technica|

#gemma-4-12b #google #local-ai

technology1 month ago•14 min saved

RTX Spark Aims to Put AI on Your Desktop, but Memory Shortages Could Slow It

Nvidia and Microsoft unveiled RTX Spark, a Windows PC concept that combines a 20-core Arm CPU (Grace lineage) with a Blackwell-based RTX GPU to run large AI models locally on consumer devices, aiming to cut latency, privacy concerns, and cloud costs while spurring a new PC upgrade cycle. However, memory constraints, including tight supply and rising costs for high-bandwidth memory (HBM), could impede adoption and complicate the economics relative to premium incumbents like Apple.

via Yahoo Finance|

#local-ai #memory-shortage #microsoft

technology2 months ago•5 min saved

Gemma 4 triples on-device AI speed with speculative token drafting

Google's Gemma 4 open models gain a 3x speed boost on local hardware through Multi-Token Prediction (MTP) drafters that guess future tokens and verify them in parallel with the main model, using shared memory and sparse decoding to accelerate generation without sacrificing quality. The approach works across devices from Pixel phones to Apple M4, with varying gains by hardware and under Apache 2.0 licensing.

via Ars Technica|

#edge-ai #gemma-4 #local-ai

technology2 months ago•3 min saved

Mac Mini Becomes Local AI Infrastructure Amid Memory Shortages

Apple's Mac Mini is increasingly serving as local AI infrastructure as high-memory configurations become scarce; demand from developers for on-device inference is outpacing supply, pushing 32GB/64GB Minis and Mac Studio builds into longer delays while lower-memory models remain more available. This trend mirrors a broader shift toward local AI workloads and may foreshadow a refresh cycle for Apple's desktop lineup.

via TechSpot|

#apple #local-ai #mac-mini

technology5 months ago•8 min saved

Trying a free, local coding AI stack: Goose, Ollama, and Qwen3-coder

A ZDNET tester explores a fully local, free coding AI stack using Goose (agent framework), Ollama (LLM server), and Qwen3-coder, detailing installation, a sample WordPress plugin test, and notes that while the local setup runs on a powerful Mac with 128GB RAM and can be competitive with cloud options, early results show accuracy issues and multiple retries; it's promising but not yet ready to fully replace Claude Code or Codex.

via ZDNET|

#coding #goose #local-ai

technology5 months ago•8 min saved

DIY Local AI Coding Stack: Goose, Ollama, and Qwen3-coder Run Offline on Mac

A ZDNet piece tests a fully free, offline coding AI stack built from Goose (an open‑source agent framework), Ollama (an LLM server), and the Qwen3-coder model to compete with Claude Code. The setup runs on a Mac (with 128GB RAM) using a 17GB Qwen3-coder:30b model and a 32K context, avoiding cloud sign‑ins. After installing Ollama, exposing it to the network, and configuring Goose to use the Qwen3-coder model, the author tests a simple WordPress plugin: results improve after several iterations but are not flawless yet. The article notes this local combo can approach paid options in some respects, but remains early in its development, with further deep dives and larger-project tests promised in upcoming installments.

via ZDNET|

#coding-tools #goose #local-ai

technology5 months ago•9 min saved

Clawdbot Comes to Mac Minis, but Privacy Concerns Outweigh Convenience

Clawdbot is a locally run AI assistant that can operate via chat apps on Mac, Linux, and Windows—often deployed on Mac minis—where it can automate tasks across apps and remember past interactions. While powerful, it raises significant privacy and security risks due to full device access and potential prompt-injection, and its value may not justify the hardware and risk; setup is technical with guidance available via its GitHub.

via Lifehacker|

#ai #local-ai #mac-mini

technology10 months ago•3 min saved

Ecovacs Unveils Fast-Charging Deebot X11 OmniCyclone at IFA 2025

Ecovacs' new Deebot X11 OmniCyclone robot vacuum features mostly on-device AI for identifying messes and adapting cleaning routines, reducing reliance on cloud connectivity, though full functionality still depends on internet access for app control and voice assistant features. The move towards local AI is promising for privacy and reliability, but its success depends on the effectiveness of onboard AI capabilities.

via Gizmodo|

#ai #ecovacs #local-ai

technology11 months ago•7 min saved

OpenAI's Latest Models and Innovations in AI Accessibility

The article discusses OpenAI's 'gpt-oss', an open-weight AI model that can be run locally on personal devices like Macs, offering privacy benefits but suffering from slow performance compared to cloud-based models like ChatGPT. The author tested it on two Macs, finding it slow but private, and highlights its potential for users with powerful hardware who prioritize privacy over speed.

via Lifehacker|

#gpt-oss #local-ai #mac

technology11 months ago•2 min saved

Building a Gaming PC to Run AI Models: My Experience

The article discusses building a high-end gaming PC to run powerful local AI models, highlighting the advantages of privacy and independence from internet connectivity, and detailing the hardware specifications and costs involved.

via CNET|

#ai-models #gaming-pc #hardware

technology11 months ago•71 min saved

Creating a Local Offline AI Workspace

The article discusses the challenges and future prospects of building a fully offline, self-hosted AI workspace, emphasizing hardware limitations like RAM and processing power, the importance of efficiency, and the potential of emerging hardware and software solutions to make local AI more accessible and practical in the coming years.

via Hacker News|

#llm-hardware #local-ai #privacy

technology1 year ago•2 min saved

Google Launches New Tools for Faster, Offline On-Device AI

Google has released the AI Edge Gallery app, allowing users to download and run AI models locally on their phones without internet, supporting tasks like image generation, question answering, and coding, with a focus on privacy and offline use.

via TechCrunch|

#ai-edge-gallery #ai-models #google

technology2 years ago•5 min saved

"Nvidia's Free Chat with RTX AI: Run GenAI Models on Your PC"

Nvidia has introduced Chat with RTX, a local AI chatbot that allows users to utilize an AI model to browse through offline data using retrieval-augmented generation (RAG). The guide explains how to download and use the tool, including adding and refreshing datasets, selecting AI models, and using it with YouTube videos. The tool requires an RTX 40-series or 30-series GPU with at least 8GB of VRAM, 16GB of system RAM, 100GB of disk space, and Windows 11, and may encounter installation issues.

via Digital Trends|

#ai-chatbot #chat-with-rtx #local-ai