NVIDIA said RTX Spark laptops are available from Friday, October 16, following an announcement with Microsoft at an event in San Francisco a week earlier.
What the chip offers
NVIDIA describes RTX Spark as a single superchip combining gaming, creative and AI-computing capabilities, with up to 128GB of unified memory and 1 petaflop of AI compute. The pitch is running personal AI agents locally rather than in the cloud.
Software around the launch
NVIDIA lists several launch-week updates:
- Nemotron 3.5 Bolt 74B, a model made for agentic AI on RTX Spark, was due on October 15.
- Agent apps including GitHub Copilot, Perplexity and others are updating for easier setup, with security via OpenShell and Microsoft Execution Containers and hybrid local and cloud routing.
- PyTorch, Python, CUDA and its libraries are available, and llama.cpp adds a Windows app.
- Adobe says up to 2x performance is coming soon to its top apps, per NVIDIA.
These are vendor statements; independent performance testing has yet to appear.
Why it matters
Large unified memory is what lets a laptop hold models of tens of billions of parameters, so this launch tests how far local agents can go on consumer hardware.