Thin, Light and Built for Everything: NVIDIA RTX Spark Windows Laptops Gain New Models, Agent Tools and Games
Source
blogs.nvidia.com
Author
Shreevar Rastogi
Date
Why it matters
Local inference hardware with 128GB unified memory in a laptop changes which models you can run on-device. The new GPU split between foreground apps and background agents targets local AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → workloads.
Key takeaways · AI-distilled
RTX Spark laptops, now shipping from ASUS, Dell, HP, Lenovo, Microsoft and MSI from $2,599, put a Blackwell RTX GPU and Grace CPU on one superchip with up to 128GB of unified memory and up to 1 petaflop of FP4, so models and creative work share one memory pool.
Nemotron 3.5 Transcribe is one open weightsA model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.Full definition →, promptable model for transcription, speaker labels, translation and audio questions in 40 languages. NVIDIA says it ranks No. 1 on Hugging Face Open ASR with about 4.3B active parameters, 80% fewer than the top open rival.
A new NVIDIA App setting lets users choose whether the GPU favors the foreground app or a long-running headless agentAn agent running with no one watching — in CI, on a schedule, or triggered by an event — with its output reviewed after the fact.Full definition →, under System > Advanced; it requires Windows hardware-accelerated GPU scheduling to be enabled.
Liquid AI's LFM3 family spans text, vision, audio, omni and decision models on one shared backbone, with open weights; its new 1.7B base model outperforms the previous 2.6B base, and it runs on RTX Spark through llama.cpp and vLLM with local fine-tuningTaking a trained model and training it a bit more on your own examples so it gets better at one specific job.Full definition →.
Terms in this piece · Glossary
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
headless agent — An agent running with no one watching — in CI, on a schedule, or triggered by an event — with its output reviewed after the fact.
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
