Meet Alpamayo 2 Super, now commercially available for robotaxis and autonomous vehicles. Built for complex real-world driving, this open reasoning model adds 360° awareness, high-level driving decisions and automated reasoning labels. We built it, so you can build on it.

Read the blog →

Available on @huggingface :

The demo is the easy part. The real test is rain, bad lane paint, and one confused driver doing something stupid. How much of that is in the benchmark?

wild it's open weight though, nvidia basically handing tesla and waymo's homework to everyone else for free. 32B params, triples the last version, and it's threatening companies that spent years building this in house. classic nvidia move, sell the shovels then also open source a free shovel factory

Nivdia Salam 3CZR212rNfJFjiWdt4eaNV5wJ9D1JxVNZMqqVGoxpump Ape this for culture as they spreading peace and good shi. 35% locked Spread good lessons and good teachings Also all the fees go to support mosque, education for underprivileged and other good things

The future of civilization is autonomy.

@DavidMoss and I would love to come test this out and make a video.

We built it, so you can build on it… and I already called shotgun on the first Alpamayo-powered robotaxi. 360° awareness + open reasoning model that actually thinks? NVIDIA just handed the entire AV industry the keys and said “go faster.” This is how you accelerate the future.🫡

Elon just said he will only buy Nvidia. But now Jensen is selling his own version of Tesla FSD?

The next revolution in transportation is not about cars that drive themselves. It’s about machines that can understand.

Let people install it in their personal cars!

If Kimi K3 is the ultimate open server for text, Alpamayo 2 Super is the ultimate open engine for autonomous driving. When a text model hallucinates a word, it's just a glitch. When a car model hallucinates a stop sign, it's a fatal crash. For that, Nvidia deserves applause.

Awesome — open reasoning models with 360° awareness for real-world robotaxis? Huge step. Can’t wait to see teams building on this. 🚀

Previously I notticed the model named "Locate Anything", which open sourced in hugging face. This Alpaymayo 2 also working with the same tech and it have some personal decision making skill on it's own, thats the only change. But it's a huge step on open autonomous driving tech.

@huggingface 🔥🔥🔥

I want to be able to get this out in a computer physics simulation - is it open source for that too?

Open source is genuinely getting amazing now. With all of these Chinese coding models, and now even more open source models that would be genuinely amazing to use is great. So thankful that people aren’t gatekeeping these amazing software tools.

Some Tesla investors got spooked by this announcement. It does empower Tesla’s competitors who’ve been lagging behind by miles. But this is a foundation model, one layer of the autonomous solution. You still need to use real world and simulation data training. It’s always good to have competition and if another solution can save those millions got killed by human driving errors, then it’s a good thing for the society overall. Tesla is leading far ahead, with 13B real life high quality marked data, vertically integrated product line.

“Great work on Alpamayo 2 Super, sir. Curious about how backside, left, and right‑side awareness are handled since 360° awareness is critical for real‑world driving

The interesting product shift is not just “reasoning.” It’s evaluating the consequences of a sequence of actions. For physical AI, closed-loop tests expose compounding errors that a static benchmark can hide. That’s the bar I’d watch.

automated reasoning labels are interesting tho curious how the inference latency holds up when da system needs that 360 awareness in real time

Good thing no one will use this trash

Open reasoning models are helping accelerate autonomous mobility.Providing developers with a commercially available foundation for perception, reasoning, and decision-making can drive faster innovation across autonomous vehicles and robotics. 🚗🤖

commercial open weights for L4 robotaxis will speed up autonomy by years

The biggest leap isn't just perception it's reasoning. Seeing the road is one thing; making safe decisions in unpredictable situations is another. What real-world driving scenario do you think is still the hardest for AI to solve?

This is amazing!!!

Can I use it to drive my own car as a courier driver?

0x7e6D8536Ae9082e380BC943763e1274511561E18

世界會因自動駕駛變得更安全

Alpamayo 2 Super’s reasoning is a game changer. Regarding OpenMDW-1.1: How does it quantify reasoning vs. modular stacks in edge cases? Also, does the license cover liability frameworks for commercial deployment? Safety guardrails are as crucial as raw performance for mass adoption.