Jensen Huang, NVIDIA CEO: "It even runs large language models" — a $249 AI computer that fits on your desk.

This literally came out in December 2024, why are you posting it now?

Here's my idea. Mount the Jetson board (or a compatible carrier) inside your Dell case using 3D-printed mounts, extra PSU power, and USB/Ethernet passthrough. People do this for custom mini PCs. See below. (Examples of the Jetson board and custom mini PC cases people build around it.) Practical Ways to Run It "in a Shell" on Your Desktop This is the easier and more common approach—treat the Jetson as a networked AI companion: Remote Desktop / Shell Access Run the Jetson headless (no monitor) and connect via SSH for terminal/shell access. For full GUI: Set up VNC (x11vnc or RealVNC), RDP (xrdp), or NoMachine. Access it seamlessly from your Windows desktop like a virtual machine window.medium.comTools like Dataplicity or simple port forwarding make it feel integrated. Networked AI Workloads:Run local LLMs, image gen, or vision models on the Jetson (via Ollama, TensorRT, etc.). Call it from your main PC using APIs, Python scripts, or tools like ssh + remote execution. Great for offloading heavy AI tasks while your Dell handles everyday work. Other Integration Ideas:USB/Ethernet bridging for direct communication. Docker or containerized services on the Jetson exposed to your desktop.

Straight out of the baking tray it's a cake y'all. Everything is a cake. Don't get fooled 😅😂

Soon, the BEST open-world gaming experiences will be found by having a local AI that your gaming platform connects to. The AI will work with your game engine to drive all the NPC interactions, keeping their agency and memories straight for an ultimate role-playing experience!

But can it run DOOM?

It’s just gonna keep getting smarter and smarter and better and better and cheaper and cheap cheaper That is the miracle of exponential growth, and the law of accelerating returns, recognized by the great Ray Kurzweil

@grok when was it released? Give its specs

Examples that actually run well on this overhyped board: Llama 3.2 3B Gemma-3/4B, Qwen2.5 3B 7B class models work but are slower, ~6–7 tokens/sec. Anything bigger quickly hits the memory wall For comparison, Kimi 2.6 is 100x larger than what the Jetson can handle.

AI for ants?

Its good these accounts exist so they show their true colors of the poor quality of their posts, then we can easily mute them or block them.

I have a couple of these devices. Really cool to work with for local ai inference and real-time object detection (with models like yolo and ultralytics package) The amount of real world problems you can solve with such a small device is incredible.

Holy sh*t

jetson nano's been out for a while now, isn't it a bit old? (i used it in a few commercial projects back in 2019)

249 dollar AI compute is basically free. that's when the scarce thing flips from computing power to knowing what to run and why. the human question — what problem is worth solving — gets more valuable as the execution layer gets cheaper.

Bro at the the time didn't know how big LLMs would get.

This is insane 😂 Jetson Orin Nano Super is already available in Thailand (~12,200 THB at Cytron & RS Components). Perfect for running LLMs locally! 🔥

@grok explain what this could do, and what its purpose is

I have a Jetson Nano, early model. Seeed Studio sells the reComputer J1020 which uses a Jetson Nano 4GB Production Module in a device that is more like a complete computer. These are mostly used for computer vision and robotics, not LLMs.

this is not new but bye bye raspberry pi you are radio shack now

The fact that a $249 desktop-sized system can run large language models shows how fast AI hardware is evolving 👀 What used to require massive servers is slowly becoming accessible to everyday developers and creators.

@grok send me the link to pay this product

B/C $NVDA is pumping

$NVDA..... TO THE MOON!

...and? What is the deal here exactly? 8GB RAM? Yeah?

Can it run doom ?

time to save some funds...

@grok what are the models that are as close to par with Frontier Models can I run on this?

Needs a MAJOR update- 8GB and old processor. Need a 32GB version.

You know what I'm REALLY impressed. 67 TOPS is better than my 4080, Plus 8GB RAM + 6 Core Compute on a glorified Raspberry Pi, For $250. That's CHEAP try and buy that compute in pieces off FBM.

I really don’t understand why are we even need them? Can someone explain me please?🙏

How many gigs of vram? let me guess. 12! No, wait. 6!

He definitely cooked, invest in NVIDIA ASAP.

I have two of them! They are amazing!

Wow!!!!