NVIDIA's Nemotron 3.5 Super VL is a 120-billion-parameter model that combines reasoning, tool use and multimodal understanding. A new guide shows developers how to run it on Amazon EKS and SageMaker HyperPod. It is aimed at engineers building agentic AI systems.
How to try it
Open the AWS Machine Learning Blog post and follow its steps for Amazon EKS or SageMaker HyperPod. Start with an AWS account, since the deployment runs on Amazon's cloud.
NVIDIA's blog describes how developers now pair frontier AI models with NVIDIA Omniverse libraries to build simulations. The AI helps assemble assets, connect physics and rendering, and check that a scene behaves as intended. It is aimed at developers and simulation builders, not everyday users.
How to try it
Find Omniverse on NVIDIA's developer site and open its libraries. Start with one small scene and ask an AI agent to assemble and check it.
NVIDIA has released Nemotron 3.5 Transcribe, an open-weights speech model that transcribes 40 languages. It adds speaker labels and word timestamps, so you can see who said what and when. It has 4.3 billion active parameters.
How to try it
Find the open weights on Hugging Face and test the model on a short audio clip.
Nvidia released Nemotron 3.5 Transcribe, an open speech recognition model that handles transcription, speaker recognition and translation. You give it instructions as plain text prompts, and it leads several benchmarks, according to the story, so it suits anyone working with recorded calls or meetings.
How to try it
Look for Nemotron 3.5 Transcribe on Hugging Face, where Nvidia's open models are usually shared. Start by uploading a short audio file and typing a prompt that asks for speaker labels.
NVIDIA has added new models, agent tools and games to its RTX Spark Windows laptops. These thin and light laptops are sold by ASUS, Dell, HP, Lenovo, Microsoft and MSI. They suit developers, creators and gamers.
How to try it
These are laptops, so you would need to buy one from a listed brand. Start by checking the new models on the NVIDIA or laptop maker's website.
NVIDIA released Nemotron 3.5 Transcribe and Granary v2, multilingual speech-to-text models you can prompt and adapt. They turn spoken audio into written text across many languages.
How to try it
Look for the model on Hugging Face or NVIDIA's site and try a short audio clip first.
The US Justice Department is looking into Nvidia's planned acquisition of Hugging Face, an open-source platform popular with AI developers. Antitrust regulators are also scrutinising Nvidia's sales practices. The review is still at the examination stage.
US intelligence agencies believe a person in the inner circle of Nvidia CEO Jensen Huang is working as an informant for China. The report, cited by The New York Times, says Chinese spies targeted the chip giant's leader. The Decoder carried the story, which links back to the original New York Times account.
Nvidia has released Nemotron 3.5 Transcribe, an open-weights speech model that turns audio into text. It handles 40 languages and adds speaker labels and word timestamps. It reports a 4.71% word error rate on the English Open ASR Leaderboard, a public test ranking.
Quick glossaryNvidia, open-weights, word error rate
Nvidia
A US chipmaker best known for the graphics processors that power most AI training.
open-weights
A model whose trained settings are free for anyone to download and run on their own computers.
word error rate
A score showing how often a speech tool gets words wrong; lower numbers mean better accuracy.
The US Justice Department is reportedly questioning Nvidia about its purchase of Hugging Face. The questions cover Nvidia's acquisitions, investments and sales practices. Officials are also asking whether the chip designer has pressured its customers. The report cites two people familiar with the matter.
Quick glossaryNvidia, Justice Department, Hugging Face
Nvidia
A US chip company best known for the powerful graphics chips that train and run modern AI systems.
Justice Department
The US government department that enforces federal laws and handles legal cases, including checks on company behaviour.
Hugging Face
An online platform where developers share and download AI models and datasets, often for free.
Nvidia has released Nemotron 3.5 Transcribe, an open speech recognition model. It handles transcription, speaker recognition and translation, and it follows instructions given as text prompts. The model leads several benchmarks, which are standard tests used to compare AI models.
Quick glossaryNvidia, speech recognition, open model
Nvidia
A US chipmaker best known for the graphics processors that power most AI systems.
speech recognition
Software that turns spoken words into written text, like the voice typing on a phone.
open model
An AI model whose code and weights are released publicly so others can download and use it.
NVIDIA RTX Spark laptops are now available from ASUS, Dell, HP, Lenovo, Microsoft and MSI. The thin and light systems combine an NVIDIA Blackwell RTX GPU with an NVIDIA Grace CPU. New models, agent tools and games are part of the launch for developers, creators and gamers.
Quick glossaryNVIDIA, GPU, CPU
NVIDIA
A US chipmaker best known for the graphics processors that power most AI and gaming hardware.
GPU
A chip that handles many calculations at the same time, used for graphics, games and AI.
CPU
The main processor in a computer that runs most everyday programs and tasks.
A Network World headline covers Nvidia's GTC Berlin event and its announcements. It points to faster AI data centres, lower power use and more secure agents.
Quick glossaryNvidia, AI factories
Nvidia
A US company best known for graphics chips that power AI computers, based in California and led by Jensen Huang.
AI factories
Large data centres built to train and run AI models quickly, using powerful chips.
A Genetic Engineering News headline says Nvidia and the UW Institute for Protein Design have released an open-source protein design AI model. It means scientists can use and build on the model for free.
Quick glossaryNvidia, Open-source
Nvidia
A US company best known for graphics chips that power AI computers, based in California and led by Jensen Huang.
Open-source
Software whose code is shared publicly so anyone can use, study, change and share it.
NVIDIA Nemotron 3.5 Super VL
AWS Machine Learning Blog
