Tag

Local Ai

All articles tagged with #local ai

technology2 days ago•7 min saved

GitHub Copilot Integrates Local AI Models and Sandboxing for Secure Edge Computing

GitHub Copilot now supports on-device inference using Microsoft’s MAI Code 1.1 Flash model, enabling developers to run coding tasks locally on Windows PCs with high memory capacity. This feature, rolling out by late October 2026, automatically routes tasks between local and cloud models to optimize speed and cost. Simultaneously, local sandboxing via Microsoft Execution Containers (MXC) is generally available, securing agent actions by restricting file, network, and credential access across Windows, macOS, and Linux.

via Microsoft|

#github-copilot #local-ai #microsoft

technology2 days ago•16 min saved

Microsoft Surface Laptop Ultra Launches Oct 16 with Magnetic Charging and Local AI Focus

Microsoft’s Surface Laptop Ultra, powered by NVIDIA’s RTX Spark chip, launches on October 16, 2026. Priced from $2,600 to nearly $6,000, it features a 15-inch mini-LED display, up to 128GB unified memory, and a magnetic 140W USB-C port. The device targets creative professionals and AI developers by enabling local execution of large models, while Microsoft simultaneously rolls out 'hybrid intelligence' features for Windows 11 to manage AI agents securely.

via Gizmodo|

#gaming-laptops #local-ai #microsoft-surface

technology2 days ago•3 min saved

Microsoft Launches Surface Laptop Ultra and Local AI Features at San Francisco Event

Microsoft unveiled the Surface Laptop Ultra, a premium 15-inch device powered by Nvidia’s RTX Spark chip, alongside new local AI capabilities for Windows 11. The event, held in San Francisco on October 7, featured keynotes from Satya Nadella and Jensen Huang, emphasizing on-device AI processing and agent security. Preorders for the Ultra began immediately, with shipping scheduled for October 16, while Microsoft detailed its strategy for 'hybrid intelligence' to reduce reliance on cloud services.

via PCMag|

#local-ai #microsoft #nvidia-rtx-spark

technology3 days ago•14 min saved

Microsoft Unveils Dual-Tier Surface Laptop Ultra to Challenge Apple’s MacBook Pro

Microsoft is set to reveal the Surface Laptop Ultra on October 7, featuring Nvidia’s RTX Spark chip in two variants: an 18-core model with up to 32GB of memory and a 20-core flagship with 128GB. The device targets Apple’s MacBook Pro by supporting local 120B-parameter AI models and offering high performance on battery, with pricing expected to be steep.

via Gizmodo|

#local-ai #microsoft #nvidia-rtx-spark

hardware4 days ago•13 min saved

NVIDIA DGX Spark 64GB Launches at $4,999 Amid Memory Shortage

NVIDIA is launching a 64GB configuration of its DGX Spark mini PC on October 23, 2026, priced at $4,999. This new model, available through partners like ASUS, Acer, and Dell, offers the same GB10 Grace Blackwell Superchip as the 128GB version but with half the unified memory. While it supports up to 100-billion-parameter models, it costs $1,000 more than the 128GB model's original launch price, reflecting the ongoing global memory shortage.

via TechPowerUp|

#asus #dgx-spark #hardware

technology6 days ago•7 min saved

NVIDIA DGX Spark 64GB Launches at $4,999 as 128GB Price Surges to $6,950

NVIDIA is launching a new 64GB configuration of its DGX Spark local AI workstation on October 23, 2026, priced at $4,999. This move follows significant price hikes for the original 128GB model, which now costs approximately $6,950. The new unit retains the GB10 Grace Blackwell Superchip and full software stack but halves the unified memory, supporting up to 100-billion-parameter models. Users can cluster two 64GB units to reach 128GB total memory and support larger models, though this requires an additional purchase.

via ServeTheHome|

#ai-hardware #dgx-spark #local-ai

hardware7 days ago•20 min saved

NVIDIA Launches 64GB DGX Spark at $4,999 Amid Memory Shortage

NVIDIA is releasing a 64GB variant of its DGX Spark desktop AI system on October 23, 2026, priced at $4,999. This new configuration addresses the ongoing memory shortage by offering a lower entry point than the 128GB model, which has risen to $6,950. The 64GB unit retains the GB10 Grace Blackwell Superchip and supports up to 100-billion-parameter models, with clustering capabilities to expand capacity.

via TechPowerUp|

#ai-hardware #dgx-spark #hardware

technology7 days ago•1 min saved

NVIDIA Launches 64GB DGX Spark at $4,999 as Memory Costs Drive 128GB Model to $6,950

NVIDIA is releasing a new 64GB variant of its DGX Spark AI workstation on October 23, 2026, priced at $4,999. This move follows significant price hikes for the original 128GB model, which now exceeds $6,000 and is reported to reach $6,950. The new configuration retains the GB10 Grace Blackwell Superchip and full software stack but halves the unified memory. While the 64GB unit supports up to 100-billion-parameter models, two clustered units can pool memory to 128GB and support 200-billion-parameter models with a 1.7x performance boost. The launch is a response to rising memory costs, offering a lower entry point for developers while the 128GB version becomes significantly more expensive.

via Wccftech|

#ai-hardware #dgx-spark #local-ai

technology7 days ago•18 min saved

Nvidia Launches 64GB DGX Spark at $4,999 to Counter Memory Shortage

Nvidia has released a new 64GB configuration of its DGX Spark local AI workstation, priced at $4,999, to address rising memory costs. This new model offers half the unified memory of the original 128GB version but retains the same GB10 Grace Blackwell Superchip and software stack. The move allows developers to run up to 100-billion-parameter models locally, with the option to cluster two units for 128GB total memory. The original 128GB model has seen significant price increases, reaching nearly $7,000, due to the ongoing memory crunch.

via The Register|

#ai-hardware #dgx-spark #local-ai

hardware10 days ago•10 min saved

GMKtec EVO-X5 Pro Launches with 192GB RAM and 320B Local LLM Support

GMKtec has officially launched the EVO-X5 Pro, a high-end desktop AI workstation powered by the AMD Ryzen AI Max+ PRO 495 processor. The system features 192GB of unified LPDDR5X memory, enabling fully offline inference for large language models up to 320 billion parameters. Priced at $6,799 for the 2TB storage variant, the device targets enterprise and developer markets requiring private, low-latency AI processing without cloud dependency.

via TechPowerUp|

#amd-ryzen-ai-max #gmktec #gorgon-halo

tech10 days ago•14 min saved

Framework Sets September 30 Pre-Order Date for Ryzen AI Max+ Pro 495 Desktop

Framework announced that pre-orders for its small-form-factor desktop featuring the AMD Ryzen AI Max+ Pro 495 processor will begin on September 30, 2026. The device retains the existing chassis and motherboard design to ensure backward compatibility, though it supports up to 192 GB of LPDDR5X-8533 memory, an increase from the previous 128 GB limit. While pricing has not been disclosed, the unit is expected to cost more than the current Ryzen AI Max+ 395 model. The system continues to offer Fedora or Windows 11 pre-built options and maintains support for standard Flex ATX power supplies and various cooling solutions.

via TechPowerUp|

#amd-ryzen-ai-max-pro-495 #framework-desktop #hardware-launch

hardware12 days ago•4 min saved

WiCi One Debuts Wireless eGPU Using Wi-Fi 7 to Stream RTX 5090 Performance

California-based WiCi has launched the WiCi One, an external graphics processing unit (eGPU) that connects to laptops and tablets via Wi-Fi 7 rather than traditional cables like Thunderbolt or OCuLink. The device aims to provide desktop-class GPU power for gaming and local AI inference without physical tethering. Two models are planned: a base version with an Nvidia RTX 5060 Ti 16GB and a premium variant with an RTX 5090 32GB. While the base model is available for pre-order at $1,999, the higher-end version lacks a confirmed price. WiCi claims the system can run Cyberpunk 2077 at 4K with 80 fps on the 5060 Ti and 110 fps on the 5090, though critics question the latency and bandwidth limitations of wireless connections for real-time gaming.

via TechSpot|

#egpu #hardware #local-ai

technology13 days ago•79 min saved

M5 Ultra Mac Studio Redefines Local AI Performance, but Price and Memory Constraints Persist

The M5 Ultra Mac Studio delivers a significant leap in local AI performance, offering 70% faster token generation and 150% quicker prompt processing compared to the M3 Ultra. While it remains slower than an RTX 5090 in raw compute, its unified memory architecture allows for larger model contexts and quieter operation. However, the system’s high cost and limited RAM configurations remain barriers for mainstream adoption.

via MacStories|

#apple-silicon #hardware-review #local-ai

technology13 days ago

WiCi One Wireless eGPU Launches with RTX 5060 Ti for $1,999

California-based WiCi has introduced the WiCi One, a wireless external GPU (eGPU) enclosure that connects to host devices via Wi-Fi 7 rather than traditional cables like Thunderbolt or OCuLink. The base model features an Nvidia RTX 5060 Ti 16GB GPU, an Intel Core 3 100U CPU, and 1TB of NVMe storage, with preorders starting at $1,999 and a standard price of $2,599. A more powerful variant with an RTX 5090 32GB and 4TB storage is planned but lacks a set price. WiCi markets the device for high-framerate gaming and local AI inference, claiming it can run Cyberpunk 2077 at 4K Ultra with DLSS 4 at 80 fps. However, critics question the practicality of Wi-Fi 7 for latency-sensitive gaming, noting that the 4x4 radio configuration may limit throughput to 11.5 Gbps, and suggest the device is better suited for non-latency-sensitive AI workloads. A developer preview is expected in Q4 2026.

via Engadget|

#local-ai #nvidia-rtx-5060-ti #technology

hardware13 days ago•11 min saved

WiCi One: A Wireless eGPU Using Wi-Fi 7 for Local AI and Gaming

California-based WiCi has launched the WiCi One, a wireless external GPU (eGPU) enclosure that connects to host devices via Wi-Fi 7 rather than traditional cables like Thunderbolt or OCuLink. The device features an NVIDIA RTX 5060 Ti, an Intel Core Ultra 7 255H processor, and 4TB of NVMe storage. While marketed for high-framerate gaming, WiCi emphasizes its utility for local AI inference and 3D rendering, where latency is less critical. The RTX 5060 Ti version is priced at $2,599, with an early sign-up rate of $1,999, while a more powerful RTX 5090 variant is planned but lacks a set price. A developer preview is expected in Q4 2026.

via TechPowerUp|

#egpu #hardware #local-ai