A Windows laptop that can hold a 100B-plus parameter model in memory sounded unlikely a year ago. RTX Spark is NVIDIA's attempt. Here's the plain version.
What launched
On Oct 16, 2026, NVIDIA said RTX Spark laptops are available from ASUS, Dell, HP, Lenovo, Microsoft and MSI, starting at $2,599. Pricing and configurations vary by region.
Models named at launch: ASUS ProArt P14 and P16, Dell XPS 16, HP OmniBook Ultra and OmniBook X 14, Lenovo Yoga Pro 9n, Microsoft Surface Laptop Ultra, and MSI Prestige N16 Flip AI+ and Prestige N16. NVIDIA also mentions compact desktops.
What's inside
A single superchip combining:
- An NVIDIA Blackwell RTX GPU
- An NVIDIA Grace CPU (MediaTek collaborated on the CPU design)
- Up to 128GB of unified memory
- Up to 1 petaflop of FP4 AI performance
The product page lists two tiers:
| Top tier | Lower tier | |
|---|---|---|
| GPU cores | 6,144 | 5,120 |
| Grace CPU | 20-core | 18-core |
| Memory | up to 128GB LPDDR5X | up to 64GB LPDDR5X |
Why unified memory matters
On a normal gaming laptop, the GPU has its own small pool of VRAM, often 8–24GB. Models that don't fit spill over and slow down badly.
With unified memory, CPU and GPU share one pool. NVIDIA says RTX Spark can run models up to 120B parameters with 1M context, because the whole model fits in the 128GB. It also says the memory handles 3D scenes up to 90GB.
At launch NVIDIA highlighted Nemotron 3.5 Bolt 74B-A9B for supported configurations, and agents like GitHub Copilot, Hermes, OpenClaw and Perplexity with guided installers.
The trade-offs
- Bandwidth. NVIDIA's launch materials don't headline a memory bandwidth figure. LPDDR5X is generally slower than the GDDR7 on a desktop card like the RTX 5090, so expect bigger models to fit but run slower than on a big desktop GPU.
- Power and heat. It's a slim laptop. Long agent runs or renders will be limited by cooling.
- Price. $2,599 is the starting point. The 128GB tier will cost more.
Who it's for
- Developers who want CUDA on Windows and local models on the go.
- Creators who want both AI tools and RTX graphics in one machine.
- People who'd otherwise buy a Mac for the memory but need NVIDIA's software stack.
Comparing options? See unified memory: Mac Studio vs RTX Spark vs desktop GPU, our VRAM guide and the local AI hardware sizing sheet. Running agents locally? Read how to run coding agents safely first.
