NVIDIA RTX Spark: price, specs, and verdict for local AI (2026)
The DGX Spark chip in a laptop: 128 GB unified memory, CUDA, fall 2026. We look at what matters for running LLMs locally : memory, bandwidth, speed, price—and who it's (or isn't) the right purchase for.
Updated on 09/10/2026
Where to buy it
Buying alternative for local AI. The button configuration is the one offered for purchase. On mini PCs, reserve memory for the system and check the inference engine; macOS/MLX and CUDA are not interchangeable.
Affiliate links — BestLLMfor may earn a commission from purchases at no extra cost to you, which does not influence these independently determined recommendations. As an Amazon Associate, BestLLMfor earns from qualifying purchases.
Technical specifications
- Memory128 GB of unified CPU/GPU-shared memory (GB10)
- BandwidthLPDDR5x unified memory—SoC bandwidth, lower than the GDDR7 in a RTX 5090 (manufacturer figure not provided for laptops)
- ComputeGB10: Grace Arm CPU + Blackwell GPU, CUDA stack
- Power consumptionNot disclosed (portable format / compact desktop)
- Indicative price≈ €2,660 announced for the first configurations (press, fall 2026 — to be confirmed)
- Largest modelSame foundation as the DGX Spark: 128 GB unified memory = ~70B models loadable in Q4, MoE models like gpt-oss-120B run comfortably; speed will depend on the machines’ actual bandwidth
Who is it for?
Advantages and limitations
✓ Strengths
- 128 GB of unified memory in a laptop
- Complete CUDA stack (standard inference engines)
- Five major manufacturers announced: Dell, ASUS, MSI, HP, Lenovo
- GB10 architecture already proven on the DGX Spark
✗ Limitations
- Not yet available: fall 2026, preorders not open
- Unofficial price ($1,799 = estimate, ≈ €2,660 reported in the press)
- Unified SoC bandwidth: large models can be loaded, but are not necessarily fast
- Not a gaming GPU: it is not intended to replace a RTX 5090 for 3D rendering
Verdict
Frequently asked questions
Can the NVIDIA RTX Spark run a 70B LLM locally?
The same foundation as the DGX Spark: 128 GB unified memory = ~70B models in Q4 can be loaded, and gpt-oss-120B-type MoE models run comfortably; speed will depend on the machines’ actual bandwidth.
How much does the NVIDIA RTX Spark cost?
Plan on approximately €2,660 announced for the initial configurations (press, fall 2026 — to be confirmed) (No official NVIDIA price — the $1,799 figures cited are analysts' estimates). Prices change quickly — check the current price through the purchase links.
Who is the NVIDIA RTX Spark for?
Developers and teams that want a large LLM locally on a CUDA laptop, without a desktop workstation.
RTX Spark vs. DGX Spark: what's the difference?
The same GB10 chip (Grace Arm CPU + Blackwell GPU, 128 GB unified). The DGX Spark is the development mini-PC sold by NVIDIA; the RTX Spark is the consumer version of this chip, integrated by Dell, ASUS, MSI, HP, and Lenovo into compact laptops and desktops.
When will RTX Spark laptops be released?
NVIDIA announces fall 2026 (Computex 2026). There is no firm preorder date or detailed product sheet from manufacturers as of today: the window remains indicative.
Is the $1,799 price official?
No: this is an analyst estimate. The initial prices reported in the press are around €2,660, subject to confirmation when preorders open.
Go further
Prices in euros (€) are French market prices including VAT, checked by QuelLLM. US prices differ: the Amazon buttons show the current US price.