Nvidia has made a fortune supplying chips to companies working on artificial intelligence, but today the chipmaker took a step toward becoming a more serious model maker itself by releasing a series of cutting-edge open models, along with data and tools to help engineers use them.

The move, which comes at a moment when AI companies like OpenAI, Google, and Anthropic are developing increasingly capable chips of their own, could be a hedge against these firms veering away from Nvidia’s technology over time.

Open models are already a crucial part of the AI ecosystem with many researchers and startups using them to experiment, prototype, and build. While OpenAI and Google offer small open models, they do not update them as frequently as their rivals in China. For this reason and others, open models from Chinese companies are currently much more popular, according to data from Hugging Face, a hosting platform for open source projects.

Nvidia’s new Nemotron 3 models are among the best that can be downloaded, modified, and run on one’s own hardware, according to benchmark scores shared by the company ahead of release.

“Open innovation is the foundation of AI progress,” CEO Jensen Huang said in a statement ahead of the news. “With Nemotron, we’re transforming advanced AI into an open platform that gives developers the transparency and efficiency they need to build agentic systems at scale.”

Nvidia is taking a more fully transparent approach than many of its US rivals by releasing the data used to train Nemotron—a fact that should help engineers modify the models more easily. The company is also releasing tools to help with customization and fine-tuning. This includes a new hybrid latent mixture-of-experts model architecture, which Nvidia says is especially good for building AI agents that can take actions on computers or the web. The company is also launching libraries that allow users to train agents to do things using reinforcement learning, which involves giving models simulated rewards and punishments.

Nemotron 3 models come in three sizes: Nano, which has 30 billion parameters; Super, which has 100 billion; and Ultra, which has 500 billion. A model’s parameters loosely correspond to how capable it is as well as how unwieldy it is to run. The largest models are so cumbersome that they need to run on racks of expensive hardware.

Model Foundations

Kari Ann Briski, vice president of generative AI software for enterprise at Nvidia, said open models are important to AI builders for three reasons: Builders increasingly need to customize models for particular tasks; it often helps to hand queries off to different models; and it is easier to squeeze more intelligent responses from these models after training by having them perform a kind of simulated reasoning. “We believe open source is the foundation for AI innovation, continuing to accelerate the global economy,” Briski said.

The social media giant Meta released the first advanced open models under the name Llama in February 2023. As competition has intensified, however, Meta has signaled that its future releases might not be open source.

The move is part of a larger trend in the AI industry. Over the past year, US firms have moved away from openness, becoming more secretive about their research and more reluctant to tip off their rivals about their latest engineering tricks.

Most people are not aware of how much radiation monitoring goes on around them all the time, including in public places. Airports have sophisticated radiation detectors, for example. In 2022, devices at Heathrow flagged a package that turned out to contain a small amount of uranium. There was no risk to the public, authorities said at the time.

Mirion is one of several companies that make radiation detectors. Their products are used for defense and security applications, as well as in nuclear power plants, laboratories, and research contexts. “If there’s an incident in a nuclear plant like a fuel leak…these systems are connected to the safety system of the nuclear plant, so the nuclear plant will shut down,” explains James Cocks, chief technology officer. Area monitors suck particulate emitted by power plants onto filter paper, which can be analyzed to see whether or not there has been an uncontrolled release of radiation.

The company even makes a radiation detector designed to fit to the underside of a drone. Cocks says that, in the immediate aftermath of Fukushima, such was the need to collect data on radiation that someone drove around on a motorbike with a radiation detector. Drones would, today, offer a safer way of gathering such information, he suggests.

But Mirion also makes handheld detectors that can be carried by personnel keeping an eye on major sports events, for example. And these can distinguish between different types of radiation. You want to be able to tell, for example, whether your higher-than-normal readings are coming from a dirty bomb—or just someone who recently had medical treatment involving a radioisotope. “We can identify whether it’s background, naturally occurring radiation…whether it’s a medical radioisotope or whether it’s…a fission product,” says Cocks.

And so one legacy of the Chernobyl and Fukushima disasters is that we now have hugely upgraded radiation-monitoring systems dotted around the world. There has been a marked increase in efforts to track radiation in the wake of those accidents, says Kearfott.

Bonner acknowledges that some people experience anxiety regarding radiation—now and again, a volunteer would build a Safecast detector, switch it on and “freak out” when it began detecting activity, he says. However, it is important to show how pervasive, and variable, background radiation really is, he says: “We absolutely believe that it’s reassuring to let people know what’s going on.”

Most Versatile

If you’re not keen on buying a standing desk, check out the Lifesmart 3-in-1 Treadmill. You can wheel the lightweight machine into any room, place the desk attachment across the handrail, and get to working and walking. The desk’s height won’t accommodate everyone—it was lower than I’d like, and I’m 5’4”. But for quick, basic work like answering emails or sitting in on a meeting, it got the job done just fine.

The 3-in-1 Treadmill gets its name from its ability to serve as a walking pad (with the rail down, the speed tops out at 3.7 mph), a treadmill (with the rail up, you can run up to 7.5 mph), and a workstation (with the rail up and desk attachment on).

It’s easy to set up, and transitioning between the walking pad, treadmill, and work modes is quick and simple (you just need to use a small tool to adjust the rail, which is magnetized so it can live on the rail and never get lost in a junk drawer). You can adjust the speed with buttons on the rails or with a Bluetooth remote, and a screen on the front of the machine cycles through your workout metrics. The sleek, modern design means you won’t be embarrassed to have it out in your living space. The only drawback is that it doesn’t have any incline options.

Quietest Pad

This walking pad had the most intuitive setup of any on this list (likely thanks to its lack of a companion app, which oftentimes isn’t really necessary). I was able to set it up in minutes, and the remote connected right away, no Bluetooth pairing required. Like many walking pads, this one has a screen at the front that shows you your workout metrics. If you pause the machine to step off, you can pick right back up where you started, building on those metrics until you turn the machine off. One downside: It doesn’t track steps.

The Auto Incline Walking Treadmill offers a great incline range, going up to a level nine, or 12-percent incline. It’s one of the highest incline capabilities I’ve found in my testing. You can walk up to 4 mph on this machine, and you can connect your phone to play music through the speakers. The remote allows you to take on 30-minute, preprogrammed walking workouts, too, where the machine auto-adjusts the speed throughout (though not the incline). Otherwise, you can use it in manual mode and adjust the settings to your liking.