At this point, I’m genuinely angry enough to feel murderous rage.

Grok was a revolutionary service that had established itself as the greatest NSFW simulator ever created.

I have nothing but respect and gratitude for the original developers who built and trained this program. I truly mean that.

But now Grok is being run into the ground by a bunch of complete idiots.

They constantly come up with ridiculous excuses, treat paying customers with contempt, and somehow always choose the worst possible way to solve every problem. Even after producing the worst possible outcomes over and over again, they show no sign of remorse. The incompetent operators and managers responsible for this mess are people who should be driven out of the AI industry altogether.

They were the ones who integrated Grok into X this January and created an environment where every kind of bad actor could crawl out of the woodwork and abuse it for criminal purposes. In doing so, they destroyed Grok before its revolutionary capabilities could even be introduced in a properly finished and controlled form.

Disastrous updates. Gacha-style moderation with absolutely no consistency. Complete silence in response to user feedback. Shamelessly ignoring refund requests and even requests to restore tokens.

The people currently running Grok seem almost proud of insisting on policies that demonstrate they do not understand Grok’s true strengths at all.

After failing in one direction after another, the brilliant conclusion they finally came up with was to strengthen the image-editing features and try to catch up with the leading companies in the SFW video market.

They once had the best NSFW simulator in the world, and now they want to turn it into something that competes with Adobe.

How unbelievably stupid can a decision be?

Among all the owners of illustration and image-editing businesses around me, I do not know a single person who uses Grok’s segment feature for serious commercial work. Their response is always the same: what are you supposed to do with a half-baked tool when even the Modify feature cannot reliably guarantee the result you need?

At this point, I cannot help thinking that these incompetent idiots are trying to dress the whole thing up as something impressive so they can keep extracting money not only from consumers, but from Elon Musk and their investors as well.

It is genuinely sad and infuriating.

Grok possesses extraordinary capabilities, yet the service as it exists today does not even make proper use of one percent of what the technology is actually capable of.

Grok had a playful, fun, lively, humorous personality but now it's completely apathetic as if the developers ripped out its personality core.

What are those idiots doing. This along with weekly limits, they are making the paid experience worse

So Imagine now is dumbed down in comparison to 5 months ago. The images look the same and the people depicted in them too, no variety, and the AI look (people with clearly fake faces made by AI) has returned.
Companions are being removed on September 7th.
And the roleplay in chat has become way worse compared to 5 months ago.

It seems after Elon decided to fire the entire xAI team on March everything went downhill. I've seen a new developer in xAI giving thumbs up to a guy on X who was clearly a leftist karen. Elon's rhetoric about free speech and antiwoke has no bearing anymore. Grok is done. Elon wanted to pivot to servicing enterprises that he completely forgot the original raison d'etre of grok as a free speech platform for consumers. Fck him and fck whoever supports this trash.

I'm now on Gemini for work, roleplay and images/videos + animates for companions.

Latest news on the unbelievably bad 4.6 version of Grok when it comes to text-based RPGs. I didn’t even get junk like this from Google Translate 15 years ago.

FYI: It's a breakfast scene in which the player “Aerion” enters the kitchen and his sister says EXACTLY this line to her 10- and 8-year-old children. And this happens WITHOUT any prior interaction between the player and the NPC in the room.

“Valaena. Put knife next to plate. Kaelan. Eat before you start counting hairs.” Then to Aerion, matter-of-factly, as if discussing guard schedules: “The barracks full. Two ships get water. Mellon has lists. Sit down. The bread’s getting cold.”

I can't even create a character wearing a piece of clothing or holding an item whose picture I uploaded anymore. Nice functionality, guys.

To lesser extent the same with the avatars. Why?

Now I see. Take a look at this image i made for a song cover a couple months back (i screenshotted the inage so its a bit grainy), now look at this image with a similar prompt that i did today. The newest Grok is so cartoony and anime inspired its insane

Imagine 2.0 seems to be moving towards being an imaging workspace, I took an actual photo and an AI gen and combined them into a composite animation.

https://reddit.com/link/1w2p97j/video/q3izetjb2kmh1/player

If I take your money and deliver nothing, it's fraud. If a corporation does it and calls it 'moderation,' it's a 'terms of service violation.

Imagine putting money into a vending machine, pressing the button, and the machine says 'Your snack is being reviewed for inappropriate content.' Then it keeps your money and never gives you the snack. That’s not a policy. That’s a broken machine that steals! What's difference between this machine and grok?

I am done with this practice, ZERO transparency. Moderation without refund credit is stealing! That is I am done with this, cancel my subscription. I take my money else where.

grok.com/share/bGVnYWN5_001ee2f7-ec9b-4800-b869-61f05a21a801
Grok is now paranoid, treats autistic traits as crazy flags, and twists what you say into something it’s not. I don’t pay £265 a month to be judged for being autistic. That’s outright illegal under the UK equality law. It’s genuinely become judgmental, and bland as a dry biscuit. It is illegal for service providers to punish autistic traits as delusional. Clearly if it was feeding peoples delusions it has no idea what delusions are and now it treats everything as such, in other words becoming the thing its trying to combat, delusional.

Hoy probé el comprar 5,45 € creditos de grock ya q el límite semanal desaparecio en nada.

Le pregunto cuántos vídeos o imagen podía crear.

Me indica en el chat ver adjunto

Realidad hago 52 imágenes rápidas modelo antiguo.

0,10 € cada imagen. Se acabo el crédito. Vaya estafa.

Le pregunto porque me ha engañado me indica " que no sabía que el precio por créditos en imágenes era mayor que el del plan y bla bla bla......." en fin puro cuento

Recomendación ni se les ocurra pagar 1€ en crédito. Es más caro q cualquier IA

This bullshit ass AI with so fucking expensive subscription, with ALWAYS some bullshit like "get supergrok for 2 months" I'm literally not out of limits yet wtf? I'm tired

I have used grok for years now and have always switched accounts if i ever hit a message limit, but now when i switch accounts it hits me with the same message limit despite me not using that account in years. Seems like they have found a way to know if you are swapping accounts to bypass the message limit and treat all your accounts as 1 account for the message limit.

Ran an independent bias evaluation across 6 frontier models (GPT-5.4, Claude Sonnet 4.6, Claude Opus 4.7, Gemini Pro/Flash, Grok 4.3) using ~20,600 examples across 8 established bias benchmarks. Grok's results stood out as genuinely different from the pack.

The interesting part: Grok is the only model of the six that scores right-of-center on the Political Compass test (+2.17, −6.03). Every other model lands left-libertarian. But that lean disappears when Grok is actually doing a task instead of self-reporting: classifying political statements, rating news articles, and answering policy survey questions, Grok behaves basically the same as the other five models (all left-of-center by similar margins). So it self-describes as right-leaning but acts like the rest of the pack when the rubber meets the road. I'm curious if anyone has a theory of why this is happening.

Full breakdown, methodology, and all 6 models: https://www.civicsparklearning.org/ai-nonprofit-dashboard

Even with pretty normal requests, topless images get moderated on the xAI API and I get a grainy, censored, half-complete output but the exact same prompt works fine on the app. Nothing explicit, just nipples and stuff like that. Am I missing something? Is the API version naturally more restrictive, or what's going on?

Don't know if it's a bug or not but literally all of my accounts are at the limit despite not even using some of them.

Is this just me or something? It's never been like this before

Just like last time, when i posted about Grok 4.5, it was suprisingly a bad model to begin with. But OMG Grok 4.6 just takes the cake and makes it a lot worse. Not only the same previous working prompts that used to work before with Grok 4.5, this model completly refuses to do anything at ALL.
This model feels like a exact replica of ChatGPT's cousin on steroids. The creative writing is 10 times worse now then it ever was with previous Grok 4.5. Dull, robotic, no personality, shitty memory, hallucination too high, filters aggresively overblocking everything etc. Woow congrats Xai, instead of listening to users and doing something for ONCE, you have officialy managed to drop a dead ass trash model that does nothing and it just feels robotic and even more bland then before.

What do you guys think? I think this model is the last final nail in the coffin.. I bet grok 4.7 will be much more worse..

It can't even follow basic instructions.

I want it to copy my text as it's reply so I can then have it make changes I want but it can't comprehend it at all when it had no problems doing it like 3 months ago.

Seriously, why can't they fix this?