NVIDIA RTX Spark laptops from six major OEMs go on sale starting at $2,599

NVIDIA's RTX Spark laptops are now on sale, with launch machines from Microsoft, ASUS, Dell, HP, Lenovo and MSI hitting shelves on October 16, 2026, according to a report on dev.to from ToolPilot Pro. Prices start at $2,599, and the pitch is direct: enough memory and AI compute in a portable machine to run large models locally rather than paying a cloud provider by the token.

The platform behind the name

RTX Spark is not a single laptop but the chip platform inside them — described in the report as NVIDIA's first complete client PC platform. The N1X combines a Grace CPU with up to 20 cores and a Blackwell GPU with up to 6,144 cores, up to 128GB of unified memory and a claimed 1 petaflop of FP4 AI compute. The goal is a thin notebook that can host large models on the machine itself, with no per-token charges and no data leaving the desk.

Prices by manufacturer

The dev.to report lists launch pricing as follows:

• Microsoft Surface Laptop Ultra: $2,599.99–$5,899.99
• ASUS ProArt P14: $2,599.99; ProArt P16: $2,799.99
• HP OmniBook X 14: $2,999.99; OmniBook Ultra 16: $3,199.99
• Dell XPS 16: $3,799.99
• MSI Prestige N16 Flip AI+: $3,299.99
• Microsoft Surface RTX Spark Dev Box: $5,999, arriving in November

Lenovo's Yoga 9n has no listed price yet, and devices from Acer and Gigabyte are expected later.

What early reviews measured

The report highlights the first independent benchmark numbers as the most useful part of launch day. One lab measured a 35-billion-parameter coding model at 58–89 tokens per second, while another recorded 63 tokens per second on a 120-billion-parameter model. Gaming tests showed AAA titles running above 100 FPS with frame generation enabled.

The caveats are real

Early testing was not uniformly positive. One review unit reportedly suffered a thermal failure that repeatedly cut power. Microsoft's own benchmark shows a 170-second wait before the first word of output on a 128,000-token prompt — a long cold start for long contexts. Battery life under AI load came in at roughly an hour in one early test.

Software lands alongside the hardware

The launch is timed with NVIDIA's software stack. Nemotron 3.5 Bolt model weights went live the day before, and the report frames Bolt as the model RTX Spark was built for. PAIR has left beta, a faster DLSS 5 model ships on launch day, and the developer toolchain — including PyTorch-CUDA on Windows on Arm, Python 3.15 and Conda — is available from day one.

Who should consider one

The report's guidance is blunt: buy one if you run local models daily, do AI-assisted coding, or need data to stay on your machine. Skip it if you simply want a good laptop, since a MacBook or a conventional Windows machine does everything else for less money — the local-AI advantage only matters if you actually run models locally.

Why it matters

RTX Spark opens a distinct hardware category: laptops designed first for local AI workloads rather than treating neural acceleration as a secondary feature. With up to 128GB of unified memory and petaflop-class FP4 compute on paper, these machines can plausibly host models that previously required a workstation or a hosted endpoint. For developers, that shifts the cost model from recurring API spend to a one-time purchase and keeps sensitive data on-device. The first-generation caveats — thermals, long-context latency and roughly an hour of battery under load — set the bar for what OEMs must fix in the next round. The signal to watch now is whether the day-one software stack keeps local models competitive with cloud alternatives, and whether more OEMs commit to the platform once Acer and Gigabyte arrive.