DGX Spark alternatives in 2026: the same chip, half the price, or more bandwidth
The DGX Spark made 128 GB of unified memory with CUDA a desktop product. It also made a market: four other brands sell the identical GB10 chip in their own cases, AMD's Strix Halo desktops offer the same memory for half the money on x86, and Apple sells the same amount with three times the bandwidth. This page lines them up with live prices.
Prices are the cheapest live eBay listing that passes our filters, item plus shipping, refreshed every 15 minutes. Speed is our estimate for Qwen3.8 27B · Q4_K_M with 8K of context.
What a DGX Spark is
A GB10 Grace Blackwell chip, 128 GB of LPDDR5X at 273 GB/s, 4 TB of storage and NVIDIA's software stack in a box the size of a hardback. The bandwidth is the catch: it is a quarter of a 3090's, so a 70B model fits comfortably and runs slowly.
NVIDIA DGX Spark
$7,600$59/GB
128 GB · 273 GB/s · ~11 tok/s
128 GB of unified memory with the full CUDA stack in a box the size of a paperback. Bandwidth is the limit: 273 GB/s means a 70B model at Q4 fits but talks at a few tokens per second. Two units link over 200 GbE for 256 GB.
The same chip from other brands
Identical silicon and memory; the differences are storage, case, warranty and, most of all, price on the day. Everything that runs on a Spark runs on these.
ASUS Ascent GX10
$6,015.41$46/GB
128 GB · 273 GB/s · ~11 tok/s
The same GB10 chip and 128 GB as the DGX Spark, with 1 TB of storage and a lower list price. Everything that runs on a Spark runs here.
HP ZGX Nano AI Station
$7,028.73$54/GB
128 GB · 273 GB/s · ~11 tok/s
HP's GB10 box, sold through the Z workstation channel with HP support. Same 128 GB and 273 GB/s as the DGX Spark.
Acer Veriton GN100
$6,367.19$49/GB
128 GB · 273 GB/s · ~11 tok/s
Acer's GB10 box: the same chip, 128 GB and 273 GB/s as the DGX Spark, with 2 TB of storage. Everything that runs on a Spark runs here.
Gigabyte AI TOP ATOM
$7,404.08$57/GB
128 GB · 273 GB/s · ~11 tok/s
Gigabyte's GB10 box with 4 TB of storage. Same 128 GB and 273 GB/s as the DGX Spark.
Strix Halo: x86 and half the price
AMD's Ryzen AI Max+ 395 puts 128 GB of unified memory in an ordinary PC. About 96 GB can be given to the GPU on Linux. No CUDA, slightly less bandwidth, a normal operating system and a much smaller bill.
Framework Desktop, Ryzen AI Max+ 395, 128 GB
$3,464.65$27/GB
128 GB · 256 GB/s · ~7.8 tok/s
128 GB of unified memory on x86 for half the price of a DGX Spark. About 96 GB can be handed to the GPU on Linux, enough for a 70B model at Q6. Bandwidth matches the Spark's, so speed does too.
GMKtec EVO-X2, Ryzen AI Max+ 395, 128 GB
$3,649.99$28/GB
128 GB · 256 GB/s · ~7.8 tok/s
The same Strix Halo chip and 128 GB as the Framework Desktop in a smaller box that often sells for less. Watch the cooling at full load.
Apple
More bandwidth than any of the above and the quietest machine on this page; the trade-offs are on the Mac page.
Apple Mac Studio M4 Max, 128 GB
$4,899$38/GB
128 GB · 546 GB/s · ~20 tok/s
The same 128 GB as a DGX Spark with twice the bandwidth, in a quiet desktop that is also a Mac. A 70B model at Q4 runs at a usable pace; no CUDA.
Apple Mac Studio M3 Ultra, 256 GB
$8,049.99$31/GB
256 GB · 819 GB/s · ~30 tok/s
256 GB at 819 GB/s: a 235B mixture-of-experts model at Q4, or a 70B at FP16, in a silent box. Half the price of the 512 GB configuration.
Or two used 3090s
48 GB across two cards at 936 GB/s each, bridged with NVLink, runs a 70B model at Q4 noticeably faster than any 128 GB box, for less money. You give up the models that need more than 48 GB and you gain a space heater.
NVIDIA GeForce RTX 3090
$1,489$62/GB
24 GB · 936 GB/s · ~37 tok/s
The budget 24 GB card. Nearly the bandwidth of a 4090 for a fraction of the price, and two of them bridge with NVLink for a 70B model at Q4.
More guides
- Best GPUs for local AI in 2026, by budget Every budget tier from $300 up, picked by memory first, priced live.
- Best Mac for local LLMs in 2026 Mac Studio configurations compared for running models, with live prices.
- Used server GPUs for local LLMs in 2026: P40, P100, MI50, A-series and A100 The cheap-memory cards and their fine print: cooling, power, drivers.