AIToday
Large Language ModelsArs Technica AIPublished: Oct 17, 2026, 01:00 JST

Surface Laptop Ultra hits $2,599 with RTX Spark

Surface Laptop Ultra hits $2,599 with RTX Spark

3 Key Points

  1. What happened

    Microsoft's Surface Laptop Ultra, using Nvidia's RTX Spark platform, starts at $2,599 (up to $5,899). It ran a 70-billion-parameter Llama model at 6.26 tokens/sec and played Gears of War: E-Day using 46.7GiB of memory as VRAM.

  2. Why it matters

    The machine handles high-end gaming and large local AI models that would not fit on most laptops, which may appeal to power users who want one device instead of separate work and gaming machines.

WHO IT HITSThis lands on high-performance laptop buyers who currently split work and gaming between two machines, and on developers testing on-device AI who may hit software gaps in the current preview tools.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

The Surface Laptop Ultra arrives as PC shipments fell 20 percent year over year in the third quarter of 2026, a decline that spanned consumer and commercial buying. Microsoft is betting that buyers will pay a premium for a single machine that handles both development work and gaming, rather than owning a desktop for fun and a laptop for work. The starting price of $2,599 places it firmly against the MacBook Pro rather than budget alternatives, and the company is not alone in adopting the chipset: Dell's XPS 16 Creator Edition and Asus's ProArt series are also betting on the RTX Spark platform.

The hardware side is where the review finds the most to praise. The 15-inch mini-LED display, the keyboard and the large haptic trackpad all read as premium, and the magnetic USB-C charging cable breaks away when yanked. In testing, the Blackwell GPU ran Gears of War: E-Day with ray tracing and DLSS, and reported 46.7GiB of the 64GiB unified memory as available VRAM. It also handled a 70-billion-parameter model that would be a stretch for most laptops.

The open question is software. Geekbench AI could only run through a CPU path, the AI Dev Gallery's conversational models offered no GPU pathway, and an attempt to run a CUDA-accelerated Phi model failed on a missing dependency. Microsoft itself warns buyers to check app and driver compatibility before purchase, and the company may have a more definitive answer on the base 24GB model in a few months when software and drivers catch up.

FAQ
How much does the Surface Laptop Ultra cost?
The starting price is $2,599, and depending on processor and memory options the MSRP can reach $5,899.
Can it run large AI models locally?
Yes. The reviewer ran a 70-billion-parameter Llama3.3-70b model, a 43GB file, at 6.26 tokens/sec using the GPU.
What are the software limitations right now?
In the AI Dev Gallery app, conversational Phi and Mistral models are CPU-only with no GPU path, and Phi Silica access requires a developer registration.
Ars Technica AIRead Original Article

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleMeta's October 30th terms may allow AI training on news sites