Grok 4.7 is coming... and this is going to be something completely different
Initial training is already complete
SpaceXAI is now doing supplemental training with a massive amount of SpaceX company data
And that data advantage is insane
SpaceX has spent nearly 25 years solving some of the hardest real-world engineering problems humanity has ever attempted
The level of precision SpaceX operates on is insane... it's unlike anything anyone has ever seen or attempted before
Rockets, rocket engines, satellites, manufacturing, autonomous systems, launch operations, extreme reliability, engineering at the limits of physics
Now Grok gets to learn from that engineering culture and data
Grok 4.7 is already significantly better than 4.6 and is expected to exceed every model currently available
And when it comes to real-world engineering, Elon said he would be shocked if any model beats it
Grok 4.6 brought SpaceXAI back to the absolute frontier
4.7 is being trained to push beyond it
This one is going to be very special
JUST IN - EU President Ursula von der Leyen says that people's savings accounts are a problem
and that she must put this money "to the service of European companies"
says her plan to "securitize them" and "supervise them"
They're coming for your savings, Europe.
Met a German founder this week and asked him if all the stories one reads about the challenges of startups in Germany are exaggerated. "No, they're understated." Proceeded to describe spending a full day having a 90-page investment contract read to him (mandatory under German law; § 13 BeurkG) by a notary that then charged €30,000. That was for his first company. His second company, needless to say, was not incorporated in Germany.
@buildwithrajath Can't say I've noticed a difference, to be honest. I burn about a week in a day, and it's looking like that's about the same now. I don't think the difference is noticeable, but I may be wrong.
@matviy I've had exactly the same issues. Luna was 34% one-shot in, Sol medium was 74%, Sol high was around the same, and Grok 4.6 was 100%. It's a game changer, man. This was probably over about 500 linear issues.
Grok 5 is still deep in training on the massive Colossus II cluster.
Roughly 220k high-end NVIDIA GPUs are already online, with more being added.
Power draw is sitting around 1 GW and climbing.
The model is targeting the 6 to 10 trillion parameter range, which puts it in a completely different league from anything currently public.
Training has been running since at least January.
Elon recently confirmed it should arrive before the end of 2026 and will be trained on the full 25-year SpaceX engineering data set.
That combination of extreme scale plus proprietary real-world engineering data is the real story here.
This is not another incremental Grok 4.x update.
It is the much larger architectural and compute jump happening in the background while the smaller releases keep rolling out.
When it lands it should give SpaceX a serious internal edge on engineering, simulation and design problems that current models still struggle with.
Cursor is already running on the same cluster, so the jump in capability should show up there too.
No firm date yet, but the ambition on both compute and data is very real.
It is!
You need to plug Grok Heavy into Cursor, you can monitor your Grok Heavy usage from the Grok console. That usage is separate from your Cursor Ultra usage, which is $400 in API usage, so even I could make that last a month; Im not a dev but learning and 30yrs as a project manger. It really is magical.
The Ultra API allows you to use all sorts of good cheap models too. I have 4 x $200 CC (1) OAI(3) plans to cancel now n rundown which are plugged into cursor now, also.
Tokens galore.
On the ground, there's no comparison.
I've been on Claude and Codex for months, and we're finally executing with Grok.
I'm so happy, I got Grok-heavy.
You get $700 worth of tokens ($400 Cursor Ultra n $300 Grok Heavy). What's not to like?
Plus, you get $40 X premium included.
And you end up shipping it for the first time without going round in a build, QA fail loop. I did 70 linear tickets today.
Unbelievable, amazing job, well done! Can't wait for 4.7.
NVIDIA ESTÁ PONIENDO $250.000 DE GPUs EN CASAS NORMALES
Y el dueño no paga ni un centavo por el hardware.
Es una caja del tamaño de un aire acondicionado, refrigerada por líquido, sin ventiladores, montada en la pared exterior.
Dentro: 16 GPUs Nvidia Blackwell + 4 CPUs de servidor + 3TB de RAM.
Span + Nvidia + PulteGroup lo instalan gratis.
A cambio:
→ Te pagan la luz y el internet (o te cobran una tarifa plana baja)
→ Usan la capacidad eléctrica que tu casa no está aprovechando (~60%)
Ahora mismo solo 100 casas nuevas.
El plan: 80.000 nodos y 1 gigavatio de compute para 2027.
No te pagan.
Te borran la factura.
Y tu casa se convierte en un nodo de la infraestructura de IA.
Esto ya no es ciencia ficción.
@kamellperry_ Get Grok heavy, man. It's $700 worth of tokens, $400 worth with Cursor, and $300 worth of Grok heavy. I'm not kidding. Seriously, $700 worth. And it executes. I've been going around in circles the last month with Claude and Codex. I swear they've got dumber.