NVIDIA RTX Spark Laptops Get Transcription Model, GPU Controls for Agents
NVIDIA has introduced Nemotron 3.5 Transcribe, an open weight speech model, and added GPU resource controls for background AI agents on its RTX Spark Windows laptops. The update also brings Liquid AI's LFM3 models, the game Phantom Blade Zero and RTX Remix 1.6 to the platform.
- NVIDIA introduced Nemotron 3.5 Transcribe, an open weight model for transcription, speaker labels and translation in 40 languages.
- A new NVIDIA App setting lets RTX Spark users split GPU resources between foreground apps and background agents.
- Liquid AI's LFM3 open models run on RTX Spark with native llama.cpp and vLLM support.
- RTX Remix 1.6 brings path-traced Minecraft Java Edition 1.21.1 and DLSS 5 features, and Phantom Blade Zero launched with DLSS 4.5.
- RTX Spark laptops start at $2,599 from ASUS, Dell, HP, Lenovo, Microsoft and MSI.
One speech model for transcription, speakers and translation
NVIDIA Nemotron 3.5 Transcribe is a new open weight model that turns recorded conversations into usable information. According to NVIDIA, developers no longer need separate models for transcription, speaker labels and translation, because one promptable model handles all three and can also answer questions about the audio.
The model transcribes 40 languages, including seven Indic languages. Developers steer it with plain-language prompts such as "Transcribe this recording and label each speaker." Prompts can also include product terms and formatting conventions, which NVIDIA says helps with specialized vocabulary in healthcare, financial services and customer service.
NVIDIA says the model ranks No. 1 on the Hugging Face Open ASR and No. 2 among open weight models on Artificial Analysis. It uses approximately 4.3 billion active parameters, which the company says is 80% fewer than the No. 1 open weight model. It is available on Hugging Face in BF16, FP8 and NVFP4 formats, with support on RTX Spark and DGX Spark systems, Jetson AGX Thor devices and GeForce RTX 5090 GPUs. Developers can deploy NVFP4 versions through NeMo-Speech.cpp or vLLM.
Sharing the GPU between apps and agents
RTX Spark users can run long-running agents next to their regular applications, and both draw on the same GPU compute. A new resource management feature in the NVIDIA App lets users decide how that compute is split. They can prioritize the foreground app for best performance, or give the background agent more resources so it finishes work faster.
The setting sits under System > Advanced in the NVIDIA App and requires hardware-accelerated GPU scheduling to be enabled in Windows Settings.
NVIDIA also highlighted Liquid AI's LFM3, a new generation of Liquid Foundation Models for specialized tasks. The family covers text, vision, audio, omni and decision models on one shared backbone, with base and post-trained models released as open weights. According to the source, Liquid AI's new 1.7-billion-parameter base model outperforms its previous 2.6-billion-parameter base model. Developers can run LFM3 on RTX Spark with native support through llama.cpp and vLLM and fine-tune the models locally. Use cases include agents in Hermes Agent, Pi and OpenClaw that triage inboxes, manage calendars and route support tickets.
Games and modding: Phantom Blade Zero and RTX Remix 1.6
S-GAME's action role-playing game Phantom Blade Zero launched on Oct. 28 at 7 p.m. PT on Steam and the Epic Games Store. NVIDIA worked with the studio and publisher to add ray-traced effects, which bring more accurate bounced lighting, detailed reflections and realistic shadows on foliage and terrain on RTX Spark laptops and GeForce RTX GPU-powered PCs. DLSS Dynamic Multi Frame Generation offers up to 6x frame generation, or adjusts the multiplier on the fly to reach a target frame rate.
RTX Remix, NVIDIA's modding platform for remastering classic games with path tracing and DLSS, reached version 1.6 the previous week. It lets creators and players render Minecraft Java Edition 1.21.1 through RTX Remix with DLSS-accelerated path tracing. The update adds DLSS 5 with 3D-Guided Neural Rendering, up to 6x Multi Frame Generation, native RTX Spark support and agent-assisted workflows. Portal with RTX and the Half-Life 2 RTX demo now support RTX Spark as well.
A redesigned ComfyUI workflow lets modders generate physically based materials inside RTX Remix without application programming interface setup, with batch processing and presets for less technical creators.
The hardware and what it means for users
RTX Spark laptops are available from ASUS, Dell, HP, Lenovo, Microsoft and MSI, starting at $2,599, with 128GB configurations on shelves. Each system combines an NVIDIA Blackwell RTX GPU and an NVIDIA Grace CPU in one superchip with up to 128GB of unified memory and up to 1 petaflop of FP4 AI performance. NVIDIA says the laptops are as thin as 14mm and as light as three pounds, with all-day battery life.
Because creative projects and AI models draw from one memory pool, NVIDIA says a laptop can take on work that once called for a desktop workstation, even when unplugged. For AI developers, the company points to running models and agents locally on CUDA, with a path to more than 150 million NVIDIA RTX AI PCs and cloud GPUs. For creators, RTX Spark can edit camera native 4:2:2 video and render 3D scenes of up to 90GB, and Adobe Photoshop and Premiere offer native support through beta releases. For gamers, NVIDIA says hundreds of games run on the platform.
