Skip to content

Cayman AI & desk shop

AI on your device.
A desk built around it.

AI PCs, gaming machines, laptops and tablets with documented on-device capability. Plus the desks and everyday gear that make the setup work.

Browse 93 products Prefer a custom build?

Research refreshed October 10, 2026

Concept visualization of a local AI workstation and human-reviewed workflow.

75 on-device AI picks · 18 workspace / storage picks

Build your setup

Visit the general Cayman shop
34 picks

AI PCs & mini PCs

7 picks

Gaming desktops

21 picks

AI & gaming laptops

5 picks

AI tablets

8 picks

AI tech gadgets

3 picks

Desks

11 picks

Desk gadgets

4 picks

Storage

Find your next upgrade

Manufacturer sources and AI limits on every card

Showing 93 products.

ASUS Gaming desktops

ASUS ROG NUC (2025, RTX 5080 Laptop GPU)

Local GPU AI

Core Ultra 9 275HX | 32GB DDR5 | 2TB SSD | RTX 5080 Laptop GPU 16GB VRAM

Compact gaming system for CUDA development and creation. The selected Amazon offer is a 32GB / 2TB configuration; confirm seller and SKU.

RTX Tensor Cores support local CUDA AI workloads. This is the laptop GPU, not the desktop RTX 5080; model fit depends on its 16GB dedicated VRAM.

Photo: ASUS Republic of Gamers (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

MINISFORUM AI PCs & mini PCs

MINISFORUM AI X1 Pro-370 (96GB configuration)

On-device NPU

Ryzen AI 9 HX 370 | 96GB DDR5 | 2TB SSD | Radeon 890M | OCuLink

High-system-memory mini PC with an optional OCuLink eGPU path. External GPU hardware is separate.

Dedicated NPU supports compatible on-device AI apps. NPU support depends on software; it is not a substitute for an NVIDIA CUDA GPU. The 96GB figure is system RAM, not dedicated VRAM.

Photo: MINISFORUM (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

MSI AI PCs & mini PCs

MSI EdgeXpert 11SUS (GB10, 128GB)

Local GPU AI

NVIDIA GB10 | 128GB unified memory | 4TB SSD | 20-core Arm CPU

MSI GB10 option for NVIDIA-stack developers who want the selected 4TB configuration.

GB10 runs compatible NVIDIA AI workloads locally using shared unified memory. Arm CPU and DGX OS require compatible software; 128GB is not discrete GPU VRAM.

Photo: MSI (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ACEMAGIC AI PCs & mini PCs

ACEMAGIC F5A (HX 470, 32GB)

On-device NPU

Ryzen AI 9 HX 470 | Radeon 890M | 32GB LPDDR5X | 1TB SSD

A small AMD AI PC with a complete 32GB configuration alongside separate barebones offerings.

XDNA 2 NPU supports compatible on-device AI. Listed 32GB memory is onboard; choose HX 470 because the similarly named H 255 configuration is a different platform.

Photo: ACEMAGIC (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ACEMAGIC AI PCs & mini PCs

ACEMAGIC M1A PRO+ (Max+ 395, 128GB)

Shared-memory AI

Ryzen AI Max+ 395 | Radeon 8060S | 128GB unified memory | 2TB SSD

An AMD shared-memory desktop with several SSD expansion slots and a distinctive tower-style enclosure.

Radeon 8060S and XDNA 2 accelerate supported local workloads. Memory is soldered and shared, not dedicated VRAM; the Intel/RTX M1A PRO listing is a different configuration.

Photo: ACEMAGIC (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Acer Predator AI & gaming laptops

Acer Predator Helios 18 AI

Local GPU AI

Up to Intel Core Ultra 9 290HX Plus | up to NVIDIA GeForce RTX 5090 Laptop GPU | 18-inch Mini LED

A large gaming laptop with high-end graphics for local AI workloads and demanding games.

NVIDIA RTX Laptop GPU supports compatible local CUDA inference and AI-assisted creation. Check the selected GPU and dedicated VRAM; laptop power limits and cooling affect sustained performance.

Photo: Acer Predator (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Acer Predator AI & gaming laptops

Acer Predator Helios Neo 14 AI

Local GPU AI

Up to Core Ultra 9 285H | up to RTX 5070 Laptop GPU | 14.5-inch OLED

A portable gaming laptop with a discrete RTX GPU for supported local AI workloads.

NVIDIA RTX Laptop GPU supports compatible local CUDA inference and AI-assisted creation. Check the selected GPU and dedicated VRAM; laptop power limits and cooling affect sustained performance.

Photo: Acer Predator (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Acer AI & gaming laptops

Acer Swift 16 AI

On-device NPU

Core Ultra Series 3, up to X9 388H | 16-inch OLED touchscreen | SF16-71T

A 16-inch productivity laptop with local NPU support for compatible AI tasks.

Intel NPU accelerates compatible on-device features; CPU/GPU runtimes can handle other supported local workloads. Acer's 180 TOPS figure is total platform performance, not NPU performance. Memory varies by SKU.

Photo: Acer (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Acer AI & gaming laptops

Acer Swift Edge 14 AI

On-device NPU

Core Ultra Series 2 | up to 48 NPU TOPS | 14-inch OLED | configuration varies

A lightweight Copilot+ laptop for mobile productivity and supported on-device features.

Intel AI Boost NPU supports compatible local Copilot+ features. Up to 48 NPU TOPS differs from 120 total platform TOPS; app, OS and language support determine available features.

Photo: Acer (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Alienware AI & gaming laptops

Alienware 16 Area-51

Local GPU AI

Intel Core Ultra platform | discrete NVIDIA GeForce RTX graphics options vary by configuration

A performance gaming laptop with discrete graphics for supported local AI workloads.

NVIDIA RTX Laptop GPU supports compatible local CUDA inference and AI-assisted creation. Check the selected GPU and dedicated VRAM; laptop power limits and cooling affect sustained performance.

Photo: Alienware (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Alienware Gaming desktops

Alienware Area-51 Desktop (RTX 5080, 32GB)

Local GPU AI

Ryzen 9 9950X3D2 | RTX 5080 16GB GDDR7 | 32GB DDR5 | 1TB SSD

A full-size gaming tower with NVIDIA graphics for games, rendering and compatible local AI tools.

RTX 5080 supports local CUDA/Tensor workloads with 16GB dedicated VRAM. The selected desktop configuration is distinct from Area-51 laptops; model fit depends on quantization, context and runtime.

Photo: Dell / Alienware (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Apple AI tablets

Apple iPad Air (M4)

Apple on-device AI

Apple M4 | 11-inch or 13-inch Liquid Retina display | Apple Intelligence

A touch-first tablet for supported on-device Apple Intelligence features and creative workflows.

Supported Apple Intelligence features can run on-device. Some requests may use Private Cloud Compute; feature availability depends on language, region and software version, so this is not a claim that every AI task runs offline.

Photo: Apple (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Apple AI tablets

Apple iPad Pro (M5)

Apple on-device AI

Apple M5 | Apple Intelligence support | storage and memory vary by configuration

A touch-first tablet for supported on-device Apple Intelligence features and creative workflows.

Supported Apple Intelligence features can run on-device. Some requests may use Private Cloud Compute, and feature availability depends on language, region and software version; this is not a claim that every AI task is offline.

Photo: Apple (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Apple AI PCs & mini PCs

Apple Mac mini (M5 Pro, 48GB unified memory)

Shared-memory AI

Apple M5 Pro | 48GB unified-memory configuration | Neural Engine | Thunderbolt 5

Small desktop for Apple Intelligence features and compatible local workloads, with a verified 48GB unified-memory configuration.

Apple silicon supports compatible local Metal/MLX and Neural Engine workloads. CPU and GPU share memory; some Apple Intelligence features use Private Cloud Compute.

Photo: Apple (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Apple AI & gaming laptops

Apple MacBook Air (M5)

Apple on-device AI

Apple M5 | 13-inch or 15-inch laptop | Apple Intelligence

A lightweight laptop for supported on-device Apple Intelligence features and everyday work.

Supported Apple Intelligence features can run on-device. Some requests may use Private Cloud Compute; feature availability depends on language, region and software version, so this is not a claim that every AI task runs offline.

Photo: Apple (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Arduino AI tech gadgets

Arduino Portenta H7

Built-in local AI

Dual-core STM32H747 MCU | Cortex-M7 up to 480MHz plus Cortex-M4 up to 240MHz | TensorFlow Lite Micro support

Industrial-grade microcontroller board for compact embedded projects, including supported tiny on-device ML models.

Runs supported TensorFlow Lite Micro workloads on its dual-core MCU; it has no dedicated NPU and is not a general-purpose AI computer. Requires project-specific peripherals for camera input; the board itself is not a camera.

Photo: Arduino (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ASUS AI PCs & mini PCs

ASUS ExpertCenter PN55 (Ryzen AI 400)

On-device NPU

Ryzen AI 400 family | XDNA 2 NPU | dual 2.5GbE | Wi-Fi 7

A small business-oriented AMD PC with accessible internals and support for multiple displays.

Up to 55 NPU TOPS supports compatible on-device features. RAM, SSD, processor and OS depend on the selected configuration; confirm whether the listing is barebones.

Photo: ASUS (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ASUS AI PCs & mini PCs

ASUS NUC 14 Pro AI (Core Ultra Series 2)

On-device NPU

Core Ultra Series 2 / Lunar Lake | Arc graphics | Thunderbolt 4 | sub-0.6L

A compact Intel option with a microphone, speaker and fingerprint sign-in for an everyday AI-assisted desk.

Integrated NPU provides up to 48 TOPS for supported local features. Memory and processor vary by SKU; this family is distinct from the non-AI NUC 14 Pro.

Photo: ASUS (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ASUS AI PCs & mini PCs

ASUS NUC 16 Pro (Core Ultra Series 3)

On-device NPU

Core Ultra Series 3 family | Intel NPU 5 | up to Arc B390 | kit configurations vary

A newer Intel NUC family for compact business deployments and edge-AI projects.

Intel NPU accelerates compatible local workloads, with ASUS quoting up to 50 NPU TOPS. Processor, RAM and SSD vary; kits may require memory, storage and an OS.

Photo: ASUS (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Announced

ASUS AI PCs & mini PCs

ASUS ProArt GR1X Mini PC

Local GPU AI

NVIDIA RTX Spark N1X | Blackwell RTX GPU / Grace CPU | 64GB or 128GB unified-memory options announced

Compact creator system announced with high-capacity memory options for local AI workloads; final configuration and market timing still need confirmation.

NVIDIA CUDA supports compatible local GPU workloads. CPU and GPU share unified memory; Arm/Windows software support and final regional configuration need checking.

Photo: ASUS (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Preorder

ASUS AI & gaming laptops

ASUS ProArt P16 RTX Spark (128GB / 2TB)

Local GPU AI

NVIDIA RTX Spark Superchip | 128GB unified memory | 2TB SSD | 4K OLED | H7607BA-XP99T

Preorder creator laptop combining NVIDIA Grace CPU and Blackwell RTX graphics in a shared-memory platform.

CPU and GPU share 128GB unified memory; this is not separate system RAM plus dedicated VRAM. Local CUDA workloads require compatible software for its Arm-based Windows platform.

Photo: ASUS (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ASUS ROG AI tablets

ASUS ROG Flow Z13 (2025)

Shared-memory AI

AMD Ryzen AI Max platform | 13-inch 2-in-1 gaming tablet | integrated AI acceleration

A tablet-style gaming device with a Ryzen AI platform for supported local workloads.

The Ryzen AI Max platform combines CPU, integrated graphics and NPU resources for supported local AI workloads. Software support and available memory depend on configuration.

Photo: ASUS ROG (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Announced

ASUS ROG Gaming desktops

ASUS ROG G1000 Edition 20 Gaming Desktop

Local GPU AI

AMD Ryzen 9 9950X3D | NVIDIA GeForce RTX 5090 | 128GB system RAM

Announced flagship desktop with a discrete GPU for local CUDA workloads; wait for final regional configurations before buying.

ASUS announced a 128GB system-memory configuration with an RTX 5090 for local GPU workloads. The desktop is announced, not represented as in-stock; confirm final regional specifications and GPU memory before purchase.

Photo: ASUS ROG (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ASUS ROG AI & gaming laptops

ASUS ROG Zephyrus G14 (2026)

Local GPU AI

Core Ultra 9 386H | up to RTX 5070 Ti Laptop GPU | 14-inch 3K OLED

A compact gaming laptop with discrete graphics for supported local AI and creative workloads.

RTX Laptop GPU supports compatible local CUDA workloads; Intel AI Boost NPU handles supported on-device features. US GU405 page specifies up to RTX 5070 Ti. Confirm GPU, VRAM and memory for the selected SKU.

Photo: ASUS ROG (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ASUS AI & gaming laptops

ASUS Vivobook S16 (S3607QA)

On-device NPU

Snapdragon X | Copilot+ PC | 16-inch laptop

A mainstream Snapdragon laptop with supported local NPU features for everyday work.

The Snapdragon X Copilot+ configuration supports on-device NPU features. Supported tasks and performance vary by app, operating system, language and configuration; cloud features still need internet access.

Photo: ASUS (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ASUS AI & gaming laptops

ASUS Zenbook A16 (UX3607)

On-device NPU

Snapdragon X2 Elite Extreme | Copilot+ PC | 16-inch 3K OLED

A large-screen Copilot+ laptop for portable work and supported on-device AI features.

Snapdragon X2 Elite Extreme supports Copilot+ PC on-device NPU features. Supported tasks and performance vary by app, operating system, language and configuration; cloud features still need internet access.

Photo: ASUS (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Preorder

Beelink AI PCs & mini PCs

Beelink SER10 MAX (HX 470, 64GB)

On-device NPU

Ryzen AI 9 HX 470 | 64GB DDR5 | 1TB SSD | 10GbE

A newer Beelink option pairing an AMD NPU with fast wired networking for a compact development desk.

XDNA 2 NPU supports compatible on-device inference. Up to 55 NPU TOPS is distinct from 86 total platform TOPS; general local LLMs need a compatible CPU/GPU runtime.

Photo: Beelink (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Dell AI PCs & mini PCs

Dell Pro Precision with GB10 (128GB)

Local GPU AI

NVIDIA GB10 | 128GB unified memory | 2TB SSD | DGX OS 7

Dell's compact GB10 workstation, formerly branded Pro Max, for NVIDIA-stack development at a desk.

Grace Blackwell runs compatible CUDA workloads locally. Arm CPU and DGX OS require compatible software; 128GB unified memory is shared with the system.

Photo: Dell (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Dell AI & gaming laptops

Dell XPS 14 (2026)

On-device NPU

Core Ultra Series 3 family | Intel AI Boost NPU | 14-inch | DA14260

A premium 14-inch Intel laptop for portable productivity and supported local NPU workflows.

Core Ultra Series 3 includes an Intel NPU for compatible on-device AI. GPU, display and memory vary across DA14260 configurations; do not confuse this model with earlier XPS 14 generations.

Photo: Dell (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

EMEET AI tech gadgets

EMEET PIXY Dual-Camera 4K PTZ Webcam

Built-in local AI

Dual-camera system with Sony 1/2.55-inch sensor | 4K/30 fps, 1080p/60 fps | built-in AI chip | AI tracking and gesture control

Dual-camera webcam with a dedicated AI chip for real-time tracking and framing optimization.

On-device feature: EMEET says the built-in AI chip performs real-time imaging and tracking calculations; the AI-assisted camera recognizes and tracks the subject.

Photo: EMEET (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Framework AI PCs & mini PCs

Framework Desktop (Ryzen AI Max+ PRO 495, 192GB)

Shared-memory AI

Ryzen AI Max+ PRO 495 | 192GB unified memory | 2TB SSD | Fedora prebuilt

A compact, configurable workstation for developers who want substantial shared memory and replaceable standard parts.

Integrated GPU and NPU accelerate supported local workloads. 192GB is shared system memory, not dedicated VRAM; Linux runtime support and model settings determine usable capacity.

Photo: Framework (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

GEEKOM AI PCs & mini PCs

GEEKOM A9 Max (HX 470, 32GB)

On-device NPU

Ryzen AI 9 HX 470 | Radeon 890M | 32GB DDR5 | 2TB SSD

Dual 2.5GbE and USB4 suit a small AI workstation with several monitors or networked storage.

XDNA 2 NPU accelerates supported local AI features; Radeon 890M uses shared system memory. Select the HX 470 configuration rather than the older HX 370.

Photo: GEEKOM (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

GEEKOM AI PCs & mini PCs

GEEKOM A9 Mega (Max+ 395, 128GB)

Shared-memory AI

Ryzen AI Max+ 395 | Radeon 8060S | 128GB unified memory | 2TB SSD

A larger shared-memory alternative to conventional mini PCs for supported local inference and creative workloads.

Radeon 8060S and XDNA 2 provide local GPU/NPU processing. Unified memory is shared with the OS; model size and speed depend on runtime, quantization and context.

Photo: GEEKOM (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

GEEKOM AI PCs & mini PCs

GEEKOM IT15 (Core Ultra 9 285H, 32GB)

On-device NPU

Core Ultra 9 285H | Arc 140T | 32GB DDR5 | 2TB SSD

An Intel mini PC for office work, media tools and supported OpenVINO or integrated-GPU inference.

Intel AI Boost offers up to 13 NPU TOPS for supported workloads. This is below the 40-TOPS Copilot+ NPU threshold; 99 platform TOPS does not mean 99 NPU TOPS.

Photo: GEEKOM (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

GMKtec AI PCs & mini PCs

GMKtec EVO-X3 (Ryzen AI Max+ 395, 128GB / 2TB)

Shared-memory AI

AMD Ryzen AI Max+ 395 | 128GB system memory | 2TB SSD

High-capacity compact system for local inference and creator workloads; integrated graphics share system memory rather than using dedicated VRAM.

The Ryzen AI Max+ platform supports on-device CPU, NPU and integrated-GPU workloads. The 128GB figure is system memory shared with integrated graphics, not dedicated VRAM; software support and performance vary by model and runtime.

Photo: GMKtec (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Hailo AI tech gadgets

Hailo-8 M.2 AI Acceleration Module

On-device NPU

Hailo-8 neural-network processor | up to 26 TOPS | M.2 module | PCIe host interface

Low-power M.2 accelerator for adding supported neural-network inference to compatible edge hosts.

Requires a compatible host with the correct M.2 key/PCIe lane configuration, operating-system support, and HailoRT software; not a standalone computer. Hailo rates the Hailo-8 at up to 26 TOPS. Host CPU and RAM are separate and not included in this module.

Photo: Hailo (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

HP AI & gaming laptops

HP HyperX OMEN MAX 16 (32GB / 2TB)

Local GPU AI

32GB system RAM | 2TB SSD | NVIDIA GeForce RTX 5090 Laptop GPU | 24GB dedicated VRAM

Portable RTX 5090 system for local GPU workflows; the 32GB system-memory and 24GB GPU-memory figures describe separate pools.

This exact HP configuration lists 32GB system RAM and an RTX 5090 Laptop GPU with 24GB dedicated VRAM. Compatible local GPU software is required; neither memory figure predicts model speed by itself.

Photo: HP (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

HP Gaming desktops

HP HyperX OMEN MAX 45L (64GB / 2TB)

Local GPU AI

Core Ultra 9 285K | RTX 5090 32GB dedicated VRAM | 64GB DDR5 | 2TB SSD | GT24-0963t

Gaming tower with a high-memory RTX GPU for compatible local inference, generation and creative workloads.

RTX Tensor Cores accelerate compatible local CUDA workloads. Its 32GB dedicated GPU memory is separate from the 64GB system RAM; performance depends on model and runtime.

Photo: HP (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

HP AI & gaming laptops

HP OmniBook Ultra 14 (Snapdragon X2 Plus, 16GB)

On-device NPU

Snapdragon X2 Plus X2P-64-100 | Hexagon NPU | 16GB RAM | 512GB SSD

A premium 14-inch laptop family for supported on-device AI and productivity features.

Selected X2 Plus configuration has an 80-TOPS Hexagon NPU for compatible local features. Windows on Arm requires compatible apps and drivers; cloud services still need internet access.

Photo: HP (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Check stock

HP AI & gaming laptops

HP OmniBook Ultra Flip 14

On-device NPU

Core Ultra 7 | Intel NPU | 16GB RAM | 1TB SSD | 14-fh0047nr

A convertible laptop with local NPU support for compatible on-device AI features.

Intel NPU accelerates compatible on-device AI features. HP marks this 16GB/1TB SKU out of stock on the review date; verify current retailer stock and exact processor before purchase.

Photo: HP (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

HP AI PCs & mini PCs

HP ZGX Nano G1n AI Station (GB10)

Local GPU AI

NVIDIA GB10 | 128GB unified memory | 2TB or 4TB SSD options | Arm CPU

HP's compact NVIDIA AI station for local experimentation, inference and development.

Blackwell GPU supports compatible local NVIDIA AI workloads. Unified memory is not discrete VRAM; confirm Linux/Arm software support and the SSD configuration before purchase.

Photo: HP (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Lenovo AI & gaming laptops

Lenovo Legion 9i Gen 10 (18-inch)

Local GPU AI

NVIDIA GeForce RTX 5090 Laptop GPU | 24GB dedicated VRAM | system RAM varies by SKU

Large-screen gaming laptop with a high-end discrete GPU for local GPU workloads; confirm system RAM and regional configuration before purchase.

The PSREF configuration identifies 24GB of dedicated GPU VRAM. System memory is a separate configuration detail and is not inferred; compatible local GPU software is required.

Photo: Lenovo (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Lenovo Legion AI & gaming laptops

Lenovo Legion Pro 7i Gen 10 (RTX 5090, 64GB)

Local GPU AI

Core Ultra 9 275HX | RTX 5090 Laptop GPU, 24GB VRAM | 64GB RAM | 2TB SSD

A high-performance gaming laptop with discrete graphics for supported local AI workloads.

Selected 83F50019US configuration provides 24GB dedicated GPU VRAM for compatible local CUDA workloads. Laptop RTX 5090 differs from desktop RTX 5090; 64GB system RAM is separate.

Photo: Lenovo Legion (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Lenovo Gaming desktops

Lenovo Legion Tower 7i Gen 10 (RTX 5080, 32GB)

Local GPU AI

Core Ultra 7 265K | RTX 5080 16GB GDDR7 | 32GB DDR5 | 2TB SSD

A conventional gaming tower with accessible internals for a desk that handles both gaming and local AI experiments.

RTX 5080 provides 16GB dedicated VRAM for compatible CUDA workloads. Selected 90Y6001JUS configuration has 32GB system RAM; platform upgrade limits are not installed capacity.

Photo: Lenovo (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Lenovo AI PCs & mini PCs

Lenovo ThinkStation PGX (GB10, 128GB)

Local GPU AI

NVIDIA GB10 Grace Blackwell | 128GB unified memory

Compact GB10 system for local model work in NVIDIA's AI software stack; shared memory and Arm compatibility need workload-specific checks.

The GB10 platform uses 128GB unified memory shared by CPU and GPU for local AI workloads. Runtime and framework support vary on this Arm-based system; unified memory is not dedicated VRAM and does not guarantee model speed.

Photo: Lenovo (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Lenovo AI & gaming laptops

Lenovo Yoga 9i 2-in-1 Gen 10

On-device NPU

Up to Core Ultra 7 258V | 14-inch 2-in-1 OLED | Intel AI Boost NPU

A convertible laptop for supported on-device AI features and touch-first work.

Core Ultra Series 2 NPU accelerates compatible on-device Copilot+ features. Integrated graphics share system memory; processor, memory and feature support must match the selected regional listing.

Photo: Lenovo (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Lenovo AI & gaming laptops

Lenovo Yoga Slim 9i Gen 10

On-device NPU

Up to Core Ultra 7 258V | 14-inch OLED | Intel AI Boost NPU

A thin premium laptop with local NPU support for compatible AI tasks.

Core Ultra Series 2 NPU accelerates compatible on-device Copilot+ features. Integrated graphics share system memory; processor, memory and feature support must match the selected regional listing.

Photo: Lenovo (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

LG AI & gaming laptops

LG gram Pro 16 (2026)

On-device NPU

Ryzen AI 7 | XDNA NPU | 32GB LPDDR5X | 2TB SSD | 16-inch IPS

A lightweight large-screen laptop with Ryzen AI NPU support for compatible tasks.

AMD XDNA NPU supports compatible local AI features. LG's page lists inconsistent Ryzen AI 7 suffixes (445/450); confirm the processor for SKU 16Z95U-G.AU82U1. Memory is shared with integrated graphics.

Photo: LG (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Logitech Desk gadgets

Logitech MX Master 4 Wireless Mouse

Workspace accessory

Wireless productivity mouse | 8,000 DPI sensor | haptic feedback | Actions Ring shortcuts | USB-C charging

Ergonomic multi-device mouse with customizable controls for long desk sessions and faster workflows.

Workspace input accessory; app shortcuts and haptic feedback do not perform AI inference.

Photo: Logitech (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Luxonis AI tech gadgets

Luxonis OAK-D Pro PoE

On-device NPU

PoE depth-AI camera | stereo depth and RGB | up to 4 TOPS total, including 1.4 TOPS for AI | Ethernet host connection

Networked depth camera that runs supported vision inference on-device for robotics and industrial perception.

Neural-network inference runs on the camera's embedded vision processor. Requires an Ethernet-connected host running compatible DepthAI software and a PoE source or supported power arrangement; it is a camera peripheral, not a standalone general-purpose computer. Luxonis lists 4 TOPS total compute, with 1.4 TOPS for AI.

Photo: Luxonis (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Microsoft AI tablets

Microsoft Surface Pro 13-inch (12th Edition, X2 Elite)

On-device NPU

Snapdragon X2 Elite | integrated NPU | memory varies by configuration

A detachable Windows device for supported local NPU workloads and on-device AI features; verify memory and application support for the chosen SKU.

Snapdragon X2 Elite includes an NPU for supported on-device AI features. Check Windows on Arm app compatibility; cloud assistants still need internet access.

Photo: Microsoft (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

MSI Gaming desktops

MSI MEG Vision X AI 2nd (RTX 50 Series)

Local GPU AI

Up to Core Ultra 9 285K | RTX 50 Series options | touchscreen control panel | configuration varies

A gaming desktop family with front-panel controls and NVIDIA GPU options for local generation and creation.

RTX GPU acceleration supports compatible local inference and creative tools. Select GPU and VRAM carefully: family options include RTX 5070 Ti, 5080 and 5090; maximum specs are not included in every unit.

Photo: MSI (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

MSI AI & gaming laptops

MSI Raider 16 Max HX (B2WJ)

Local GPU AI

NVIDIA GeForce RTX 5090 Laptop GPU | 24GB dedicated VRAM | system RAM varies by SKU

High-end gaming laptop with a large discrete GPU for local CUDA workloads; check system memory on the selected regional configuration.

The 24GB figure is dedicated laptop-GPU VRAM, not system RAM. Local GPU inference needs compatible CUDA software; system memory varies by configuration and is intentionally not specified here.

Photo: MSI (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

NVIDIA AI tech gadgets

NVIDIA Jetson Orin Nano Super Developer Kit

Local GPU AI

Jetson Orin Nano module and carrier board | 67 INT8 TOPS | 8GB LPDDR5 shared system memory | 102GB/s memory bandwidth

Compact standalone developer kit for robotics and edge inference with NVIDIA JetPack, CUDA, and TensorRT workloads.

Developer kit with local CUDA/TensorRT inference; storage, initial software setup and peripherals are required. Its 8GB LPDDR5 is shared system memory, not dedicated VRAM. Up to 67 INT8 TOPS requires supported software and power mode.

Photo: NVIDIA (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

OpenMV AI tech gadgets

OpenMV N6 Machine-Vision Development Board

On-device NPU

STM32N657 microcontroller | integrated NPU up to 600 GOPS INT8 | OpenMV machine-vision development platform

Programmable embedded-vision platform for building small, low-power inference projects.

The STM32N657's integrated NPU runs supported quantized models on-device at up to 600 GOPS INT8. This is a development board, not a general-purpose host PC. Confirm the chosen bundle's image-sensor inclusion and interface before treating it as a complete camera; sensor configuration is not specified here.

Photo: OpenMV (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

OWC Desk gadgets

OWC Thunderbolt 5 Hub

Workspace accessory

Thunderbolt 5 host connection | three downstream Thunderbolt 5 ports | USB-A port | up to 120Gb/s Bandwidth Boost

Compact expansion hub for connecting high-speed displays, storage, and peripherals to a Thunderbolt 5 host.

Connectivity accessory only; it does not perform AI inference. Requires a compatible Thunderbolt 5 host for full link rates; peripherals and host are sold separately.

Photo: OWC (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Raspberry Pi AI tech gadgets

Raspberry Pi AI Camera

On-device NPU

Sony IMX500 intelligent vision sensor | 12.3MP | on-sensor AI processing | Raspberry Pi camera interface

Camera module that runs supported neural-network inference on its image sensor for low-latency edge vision.

The Sony IMX500 image sensor performs supported neural-network processing on the camera module. Requires a compatible Raspberry Pi host and camera interface for control and output; it is not a standalone computer.

Photo: Raspberry Pi (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Raspberry Pi AI tech gadgets

Raspberry Pi AI HAT+ 2

On-device NPU

Hailo-10H accelerator | 40 TOPS INT4 | 8GB onboard RAM | add-on board for Raspberry Pi 5

Raspberry Pi 5 add-on for supported local generative and vision inference workloads.

Requires a Raspberry Pi 5 host and compatible Raspberry Pi software; it is not a standalone computer. Hailo-10H provides up to 40 TOPS INT4 and the board has 8GB onboard RAM. This module RAM is distinct from the Pi host's system RAM.

Photo: Raspberry Pi (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Samsung AI & gaming laptops

Samsung Galaxy Book6 Ultra (RTX configurations)

Local GPU AI

Core Ultra Series 3 | RTX 5060 or 5070 Laptop GPU options | Intel NPU

A creator and gaming laptop with discrete RTX graphics and supported Galaxy AI features.

Choose a discrete RTX 5060/5070 Laptop GPU configuration for local CUDA workloads. Book6 Ultra also has integrated-graphics variants; those do not provide NVIDIA CUDA. Intel NPU supports compatible local features.

Photo: Samsung (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Samsung AI tablets

Samsung Galaxy Tab S12 Ultra (12GB / 256GB)

On-device NPU

MediaTek Dimensity 9500 / NPU 990 | 12GB RAM | 256GB storage | 14.6-inch AMOLED

Large-screen tablet with supported on-device Galaxy AI features; check individual feature, language and network requirements.

Dimensity 9500 includes an NPU for supported on-device AI. Galaxy AI features vary: many need an account or internet; do not assume every tablet feature works offline.

Photo: Samsung (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Synology Storage

Synology DiskStation DS925+ 4-Bay NAS

Storage accessory

Four-bay desktop NAS | AMD Ryzen V1500B quad-core processor | 4GB ECC memory | dual 2.5GbE ports

Network storage appliance for centralized files, backups, and project data; drives are purchased separately.

Storage appliance only; no AI compute is claimed. Drive capacity depends on separately installed drives and supported configurations.

Photo: Synology (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Apple AI PCs & mini PCs

Apple Mac mini (M4 Pro, 48GB unified)

Shared-memory AI

Apple M4 Pro | 48GB unified-memory configuration | Thunderbolt 5

Compact Mac for local development and compatible Apple-silicon AI tools. Confirm storage and CPU/GPU configuration when choosing an offer.

Apple silicon supports local Metal/MLX workloads with compatible software. Unified memory is shared; some Apple Intelligence features use cloud services.

Photo: Apple (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

BOSGAME AI PCs & mini PCs

BOSGAME M5 (Ryzen AI Max+ 395)

Shared-memory AI

Ryzen AI Max+ 395 | 128GB / 2TB configuration | Radeon 8060S

High-memory AMD system; choose the 128GB configuration to match this listing.

Integrated GPU shares system memory. Local inference requires a compatible runtime; available GPU memory is smaller than total RAM and is not dedicated VRAM.

Photo: BOSGAME (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

GMKtec AI PCs & mini PCs

GMKtec EVO-T2 (Core Ultra X7 / X9)

On-device NPU

Core Ultra X7 358H / X9 388H options | Intel Arc graphics | Intel NPU | configuration varies

Current EVO-T2 family with Intel on-device AI acceleration. Choose the exact processor, RAM and storage before buying.

Dedicated NPU supports compatible on-device AI apps. NPU support depends on software; it is not a substitute for an NVIDIA CUDA GPU.

Photo: GMKtec (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

HP AI PCs & mini PCs

HP Z2 Mini G1a (Ryzen AI Max+ PRO 395)

Shared-memory AI

Ryzen AI Max+ PRO 395 | 128GB LPDDR5X | 1TB SSD | Radeon integrated graphics

Compact HP workstation for local AI and professional software. Check runtime support for the AMD graphics stack.

Integrated GPU shares system memory. Local inference requires a compatible runtime; available GPU memory is smaller than total RAM and is not dedicated VRAM.

Photo: HP (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Preorder

Minisforum AI PCs & mini PCs

Minisforum MS-S1 MAX (Ryzen AI Max+ 395)

Shared-memory AI

Ryzen AI Max+ 395 | Radeon 8060S | 128GB configuration | dual 10GbE

Compact workstation with high memory capacity and fast networking for local development.

Integrated GPU shares system memory. Local inference requires a compatible runtime; available GPU memory is smaller than total RAM and is not dedicated VRAM.

Photo: Minisforum (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

NVIDIA AI PCs & mini PCs

NVIDIA DGX Spark (GB10 Grace Blackwell)

Local GPU AI

NVIDIA GB10 | 128GB unified memory | 20-core Arm CPU | ConnectX-7

Personal NVIDIA AI development system for local inference and model experimentation.

GB10 runs compatible NVIDIA AI workloads locally using shared unified memory. Arm CPU and DGX OS require compatible software; 128GB is not discrete GPU VRAM.

Photo: NVIDIA (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Skytech Gaming desktops

Skytech Legacy 4 (Core Ultra 9 285K, RTX 5090)

Local GPU AI

Core Ultra 9 285K | RTX 5090 32GB VRAM | 64GB DDR5 | 4TB SSD

Gaming tower with a high-memory discrete GPU for local AI and demanding creation. Match the 4TB manufacturer configuration.

NVIDIA GPU supports local CUDA inference and generation. Its 32GB dedicated VRAM is separate from 64GB system RAM; runtime and model size still matter.

Photo: Skytech Gaming (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Apple AI PCs & mini PCs

Apple Mac Studio (M4 Max, 36GB unified)

Shared-memory AI

Apple M4 Max | 14-core CPU / 32-core GPU | 36GB unified memory

Desktop Mac with a larger GPU and front-facing ports for creative and local AI workflows.

Apple silicon supports local Metal/MLX workloads with compatible software. Unified memory is shared; some Apple Intelligence features use cloud services.

Photo: Apple (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

ASUS AI PCs & mini PCs

ASUS Ascent GX10 (GB10 Grace Blackwell)

Local GPU AI

NVIDIA GB10 | 128GB unified memory | Arm CPU | NVIDIA DGX OS

Compact GB10 workstation for developers building and testing models on the NVIDIA software stack.

GB10 runs compatible NVIDIA AI workloads locally using shared unified memory. Arm CPU and DGX OS require compatible software; 128GB is not discrete GPU VRAM.

Photo: ASUS (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Corsair AI PCs & mini PCs

Corsair AI Workstation 300 (Ryzen AI Max 385)

Shared-memory AI

Ryzen AI Max 385 | Radeon 8050S | 64GB LPDDR5X | 1TB SSD

Small workstation for local AI experiments and creative work. GPU memory comes from the 64GB system pool.

Integrated GPU shares system memory. Local inference requires a compatible runtime; available GPU memory is smaller than total RAM and is not dedicated VRAM.

Photo: CORSAIR (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Beelink AI PCs & mini PCs

Beelink GTR9 Pro (Ryzen AI Max+ 395)

Shared-memory AI

Ryzen AI Max+ 395 | Radeon 8060S | 128GB LPDDR5X

High-memory AMD mini PC for local inference and graphics workloads that support its integrated GPU.

Integrated GPU shares system memory. Local inference requires a compatible runtime; available GPU memory is smaller than total RAM and is not dedicated VRAM.

Photo: Beelink (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Beelink AI PCs & mini PCs

Beelink SER9 Pro (Ryzen AI 9 HX 370, 32GB)

On-device NPU

Ryzen AI 9 HX 370 | Radeon 890M | 32GB LPDDR5X configuration | XDNA 2 NPU

NPU-equipped mini PC for supported local AI features and integrated-graphics workloads.

Dedicated NPU supports compatible on-device AI apps. NPU support depends on software; it is not a substitute for an NVIDIA CUDA GPU.

Photo: Beelink (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

GMKtec AI PCs & mini PCs

GMKtec EVO-X2 (Ryzen AI Max+ 395)

Shared-memory AI

Ryzen AI Max+ 395 | Radeon 8060S | 128GB / 2TB configuration

High-memory mini PC for compatible local AI runtimes. This shortlist selects the 128GB version.

Integrated GPU shares system memory. Local inference requires a compatible runtime; available GPU memory is smaller than total RAM and is not dedicated VRAM.

Photo: GMKtec (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Beelink AI PCs & mini PCs

Beelink GTi14 Ultra (Core Ultra 9 185H)

On-device NPU

Core Ultra 9 185H | 32GB DDR5 / 1TB configuration | Intel Arc | Intel AI Boost NPU

Mini PC with Intel AI acceleration and optional external GPU dock support. Dock and GPU are separate purchases.

Dedicated NPU supports compatible on-device AI apps. NPU support depends on software; it is not a substitute for an NVIDIA CUDA GPU. The advertised 34.5 TOPS combines CPU, GPU and NPU.

Photo: Beelink (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

Beelink AI PCs & mini PCs

Beelink SER8 (Ryzen 7 8845HS)

On-device NPU

Ryzen 7 8845HS | Radeon 780M | Ryzen AI NPU up to 16 TOPS | RAM varies by configuration

Compact Ryzen system for supported NPU features and everyday work. Select RAM and storage carefully.

Dedicated NPU supports compatible on-device AI apps. NPU support depends on software; it is not a substitute for an NVIDIA CUDA GPU. Its 16-TOPS NPU is below the 40-TOPS Copilot+ threshold.

Photo: Beelink (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

GMKtec AI PCs & mini PCs

GMKtec EVO-T1 (Core Ultra 9 285H)

On-device NPU

Core Ultra 9 285H | Intel Arc graphics | Intel AI Boost NPU up to 13 TOPS

Intel mini PC for supported on-device AI apps. RAM and storage depend on the selected offer.

Dedicated NPU supports compatible on-device AI apps. NPU support depends on software; it is not a substitute for an NVIDIA CUDA GPU. The advertised 99 TOPS is total platform performance, not NPU performance.

Photo: GMKtec (manufacturer page, opens in a new tab)

Manufacturer specs (opens in a new tab) Reviewed Oct 10, 2026

As an Amazon Associate we earn from qualifying purchases. Search links locate offers; they do not verify an exact configuration, seller, stock or Cayman delivery. Availability labels describe manufacturer information on our research date, not live Amazon stock. Photos show model families; finishes and included accessories vary by listing.

AI labels with meaning

Computers and tablets need documented local processing. A cloud chatbot shortcut is not enough. Features may still need downloads, accounts, supported languages or cloud services.

Capacity is not a benchmark

System RAM, shared GPU memory and dedicated VRAM are different. We do not promise a model size or speed from total RAM. Confirm your runtime, operating system and exact configuration.

Research before recommendations

Each active listing links to a manufacturer source and review date. Older entries with unresolved specifications stay off this shortlist. Product photos carry source credits.

Your next setup, built around you.

Custom PC builds, desk layouts and complete workstations. Plan the hardware and software together.

Talk through your setup