Local AI
GB10, DGX Spark, RTX Spark, Strix Halo and the machines that run LLMs on your desk.
-
The AI Bubble Could Burst Soon, and Data Centers May Pop It Dalio, Burry, the IMF and the Bank of England are warning. Debt is piling up, data centers are stalling, and countries are paying with power, water and air. -
RTX Spark and DGX Spark: The Hype Faded, Buyers Feel Let Down RTX Spark searches fell 56% four days after launch, and DGX Spark interest is down by two thirds since June. Weak battery life and a 74% price hike explain why. -
Ryzen AI Max+ 495 Hype Faded Fast: Buyers Moved On Google searches for the Ryzen AI Max+ PRO 495 are down 92% from their September peak. At twice the price of a 395 box, buyers moved on within days. -
Sipeed MaixStation vs DGX Spark, Strix Halo and Gorgon Halo Sipeed MaixStation: specs, what it ships with, how it compares to DGX Spark, Strix Halo and Gorgon Halo, where it wins, where it loses and how to reserve one. -
GMKtec EVO-X5 Pro: Best Ryzen AI Max+ PRO 495 Mini PC GMKtec EVO-X5 Pro compared with every Ryzen AI Max+ PRO 495 mini PC on sale: it leads on availability, 10GbE, USB4 V2, three SSD slots and AMD DASH. -
ACEMAGIC F9A PRO 495: Specs, 192GB Memory & Price Outlook ACEMAGIC F9A PRO 495 specs: a 2-liter Ryzen AI Max+ PRO 495 mini workstation with up to 192GB unified memory and OCuLink, compared with the 395 version. -
Bosgame M5 Review: Ryzen AI Max+ 395 Strix Halo Mini PC The Bosgame M5 pairs a Ryzen AI Max+ 395 with 128 GB of LPDDR5X-8000 and a 2 TB SSD, and is the cheapest 128 GB Strix Halo mini PC on sale in 2026. $2,399–$2,999 -
Corsair AI Workstation 300 Review: The 128GB Strix Halo Box Corsair AI Workstation 300 review: Ryzen AI Max+ 395 mini PC with 128GB LPDDR5X, internal PSU, quiet modes and a 2-year warranty, but a volatile price. $1,699–$4,699 -
Framework Desktop PRO 495: 192GB Specs, Release Date & Price The Framework Desktop with Ryzen AI Max+ PRO 495 packs 192GB LPDDR5X-8533 at 273 GB/s into a 4.5L mini-ITX box for local LLMs. No price or ship date yet. -
GMKtec EVO-X5 Pro: Ryzen AI Max+ PRO 495 Specs, Release Date GMKtec EVO-X5 Pro is a Ryzen AI Max+ PRO 495 mini PC with 192GB LPDDR5X-8533, 160GB as VRAM and a 300B-parameter local LLM claim. October launch, no price yet. -
Minisforum MS-S1 Max-P495: 192GB Specs, Price & Release Date Minisforum's MS-S1 Max-P495 mini PC and N5 Max-P495 AI NAS pair the Ryzen AI Max+ Pro 495 with 192GB unified memory. Confirmed specs, price teaser, unknowns. -
Minisforum MS-S1 Max Review: 128GB Strix Halo Workstation Minisforum MS-S1 Max review: Ryzen AI Max+ 395 mini workstation with 128GB LPDDR5X, dual 10GbE and USB4 V2. Fast but loud, and we do not recommend it. $2,599–$3,799 -
OneXPlayer ONEXStation Review: Ryzen AI Max+ 395 Mini PC OneXPlayer ONEXStation review: a Ryzen AI Max+ 395 mini PC with 128GB LPDDR5X, 120W tri-fan cooling and RGB, built for local LLMs and 1080p gaming. $2,999–$3,599 -
AMD Ryzen AI Halo Mini PC: Specs, Release Date & Price AMD Ryzen AI Halo, AMD's first own-brand mini PC, pairs a Ryzen AI Max+ 395 with 128GB unified memory. Pre-orders opened June 8; due in stores July 10, 2026. -
GMKtec EVO-X3: Ryzen AI Max+ 395 Mini PC With OCuLink GMKtec EVO-X3 launches with Ryzen AI Max+ 395, 128GB LPDDR5X and an OCuLink eGPU port, one of the first Strix Halo mini PCs aimed at local LLMs. -
NIMO AI Mini PC Review: 128GB Strix Halo AI Box The NIMO AI Mini PC pairs a Ryzen AI Max+ 395 with 128GB of unified LPDDR5X, an Amazon-buyable Strix Halo workstation for local LLM inference. $2,999.99 -
Claude Code on a Mini PC: AI Coding on a Compact Desktop How to run Claude Code on a mini PC: why a modest box is enough, how it compares to local LLMs on Strix Halo machines, and how to get started. -
Acer Veriton GN100 Review: GB10 AI Mini Workstation The Acer Veriton GN100 is a GB10 Grace Blackwell AI mini workstation with 128 GB unified memory and DGX OS preloaded with PyTorch, Jupyter and Ollama. $3,999 -
ASUS Ascent GX10 Review: GB10 Grace Blackwell AI Desktop ASUS Ascent GX10 review: a GB10 Grace Blackwell AI desktop with 128 GB unified memory, its 273 GB/s bandwidth limit, and how it compares to DGX Spark. $2,999–$3,999 -
ASUS ProArt RTX Spark Mini PC: Specs, Release Date & Price ASUS ProArt Mini PC with NVIDIA RTX Spark: up to 128GB unified memory, about 1 petaflop of AI, 140W cooling and a Fall 2026 release. Confirmed specs. -
Dell XPS RTX Spark Desktop: Specs, Release Date & Price Dell's XPS RTX Spark Desktop is a GB10-class AI mini PC concept from Computex 2026, adding front USB-C and an SD card reader to NVIDIA's RTX Spark platform. -
GIGABYTE AI TOP ATOM Review: The GB10 AI Desktop GIGABYTE AI TOP ATOM review: a GB10 Grace Blackwell AI mini PC with 128GB unified memory, 1 PetaFLOP FP4 and dual ConnectX-7 for local LLM work. $3,499–$3,999 -
Lenovo RTX Spark Mini PC: Specs, N1/N1x Tiers & Price Lenovo's SFF RTX Spark mini PC debuted at Computex 2026: up to 20 Grace cores, 6,144 Blackwell CUDA cores, 128GB unified memory. Specs, N1/N1x tiers, price. -
MSI EdgeXpert MS-C931 Review: 128GB GB10 AI Desktop MSI EdgeXpert MS-C931 review: a GB10 Grace Blackwell AI desktop with 128GB unified memory and vapor-chamber cooling MSI says beats the reference DGX Spark. $3,999 -
NVIDIA DGX Spark Review: GB10 Grace Blackwell AI Desktop NVIDIA DGX Spark review: a GB10 Grace Blackwell AI desktop with 128GB unified memory that runs 200B-parameter models locally, capped by 273 GB/s bandwidth. $3,999–$4,699 -
Surface RTX Spark Dev Box: Specs, Release Date & Price The Surface RTX Spark Dev Box is Microsoft's GB10-class Windows AI mini PC with 128GB unified memory and 1 petaflop of compute. Specs and release outlook. -
AMD Medusa Halo: Zen 6 and RDNA 5 for Next-Gen AI Mini PCs AMD Medusa Halo, the Strix Halo successor with Zen 6 and RDNA 5: what AMD has confirmed for next-gen AI mini PCs, and what is still leak or rumor. -
AMD Ryzen AI Max 400 Gorgon Halo: Strix Halo Refresh Guide AMD's Ryzen AI Max 400 'Gorgon Halo' refreshes Strix Halo with up to 192 GB of unified memory and higher clocks. What it means for mini PC buyers. -
Beelink GTR9 Pro Review: 128GB Strix Halo Local-LLM Mini PC Beelink GTR9 Pro review: Ryzen AI Max+ 395 with 128 GB unified memory, dual 10 GbE and quiet vapor-chamber cooling for running 70B local LLMs. $1,899–$1,999 -
Framework Desktop Review: 4.5-Liter Strix Halo Mini-ITX Framework Desktop review: Ryzen AI Max+ 395 in a 4.5-liter mini-ITX chassis with a PCIe slot, two M.2 bays and a swappable PSU, for Linux and local AI users. $1,099–$1,999 (DIY); ~$2,500 fully built -
HP Z2 Mini G1a Review: ISV-Certified Strix Halo Mini PC HP Z2 Mini G1a review: Ryzen AI Max+ PRO 395, 128 GB unified memory, ISV certification and Tier 1 OEM warranty in a 2.3 L workstation. $3,350–$6,700 -
GMKtec EVO-X2 Review: 128 GB Strix Halo Mini Workstation GMKtec EVO-X2 review: Ryzen AI Max+ 395, Radeon 8060S and up to 128 GB unified memory in a 2-liter mini PC for local 70B AI, gaming and creator work. $1,499 -
GMKtec EVO-X2: First Strix Halo Mini PC to Reach Market GMKtec EVO-X2, the first Ryzen AI Max+ 395 Strix Halo mini PC to ship: strong hardware, an immature AI software stack, and standard consumer warranty. -
Apple Mac Mini M4 Pro for Local LLMs: 64 GB, 70B Models Apple Mac Mini M4 Pro with 64 GB unified memory and 273 GB/s bandwidth runs 4-bit 70B-class LLMs locally, with mature MLX software but no CUDA. -
NPU vs iGPU vs CPU for Local LLMs on Mini PCs (2026) NPU vs iGPU vs CPU for local LLMs on 2026 AI mini PCs: what the NPU really accelerates, what runs on the iGPU and CPU, and why memory beats TOPS. -
RTX 4070 Mini PCs for Local AI: ASUS ROG NUC vs Beelink ASUS ROG NUC 14 Performance and Beelink GTi14 Ultra with EX dock put an RTX 4070 on a small desk for local AI, one inside the box and one in a dock. -
Ryzen 8845HS Mini PCs for Local LLMs: The 7B Sweet Spot How Ryzen 7 8845HS mini PCs with a Radeon 780M iGPU and 32 GB of DDR5 run 7B and 13B local LLMs, and why most users still skip the NPU. -
GMKtec EVO-X2: Running 70B Local LLMs on Strix Halo GMKtec EVO-X2 with 128 GB unified LPDDR5X, Radeon 8060S and a 50 TOPS NPU runs 70B-parameter LLMs locally at roughly 3 to 5 tokens per second.