NVIDIA RTX Spark Laptops Gain Transcription in 40 Languages

NVIDIA RTX Spark laptops gain a speech model covering 40 languages and controls for sharing GPU resources with background agents. Systems are available starting at $2,599, alongside new support for…

Technobezz
NVIDIA RTX Spark Laptops Gain Transcription in 40 Languages

Don't Miss the Good Stuff

Get tech news that matters delivered to your inbox.

NVIDIA RTX Spark Windows laptops now support a speech model that transcribes 40 languages, alongside new controls that let owners divide GPU resources between applications and background agents. Laptops are available starting at $2,599, including configurations with 128GB of unified memory.

The systems pair a Grace CPU with Blackwell RTX graphics on one superchip, offering up to 128GB of shared memory and AI performance rated at up to 1 petaflop using FP4. Buyers can choose machines from Microsoft, Lenovo and ASUS, among six manufacturers, with designs reaching as little as three pounds and 14mm thick. The earlier laptop launch carried a starting price of $2,399, below the starting figure NVIDIA gives today.

Read more: NVIDIA RTX Spark Launch Events Open on Friday

Introduced today, NVIDIA Nemotron 3.5 Transcribe handles translation, transcription and speaker identification within a single model, and can answer questions about recordings. Its language coverage includes seven Indic languages, while developers can supply prompts containing specialist terminology and preferred formatting. NVIDIA has made its open weights available through Hugging Face in NVFP4, BF16 and FP8 formats, with deployment support extending to GeForce RTX 5090 GPUs and DGX Spark systems.

The resource controls in NVIDIA App let users favor their active application or assign more GPU capacity to an agent working in the background. Users access the setting under System > Advanced, and Windows must have hardware-accelerated GPU scheduling switched on.

Liquid AI’s LFM3 family also supports RTX Spark through vLLM and llama.cpp, with developers able to fine-tune models locally. NVIDIA says the models can support agents that handle calendars and inboxes, interpret documents or process spoken requests. According to NVIDIA, Liquid AI’s latest base model, with 1.7 billion parameters, outperforms its earlier 2.6-billion-parameter version.

For players, Phantom Blade Zero is scheduled to launch tonight, October 28, at 7 p.m. PT through the Epic Games Store and Steam, featuring ray tracing and NVIDIA DLSS 4.5. NVIDIA says its Dynamic Multi Frame Generation technology supports up to 6x frame generation and can vary the multiplier to meet a selected frame-rate target.

RTX Remix 1.6, released last week, adds native RTX Spark support and lets Minecraft Java Edition 1.21.1 use path tracing accelerated by DLSS. Portal with RTX and the Half-Life 2 RTX demo are scheduled to receive those upgrades today, including DLSS 3D-Guided Neural Rendering and additional path-tracing performance options.

Share

More in News