Z Omdömen 

Bonjour, énorme arnaque.
Prélèvement sur le compte sans autorisation.
Je vous ai envoyé un mail pour régler ce problème avant la police. Merci d’y jeter un œil. Malorie

Genuinely bad. You give it instructions and it doesn't follow it. You enforce the instructions and it follows it for exact 1-2 texts max. A problem so typical with any Chinese ai.

GLM API services are very inreliable and slow.

I have used GLM 5.1 with "GLM Coding Max"plan, and it is nearly unusable due to inreliability. The connection issues have persisted several months that I have had the plan.

Also the GLM models are not comparable to Claude/OpenAI offerings.

Not recommended.

Im canceled subscribe he always charge now 🛑Oops! Your card *1657 was declined.

Your recent charge of $80.00 USD at STRIPE-Z.AI failed because of insufficient funds

BAD SERVICE, BAD AI!!!
DON'T USE THIS SERVICE.. THIS IS SCAM SCAM SCAM!!!!

I've been wasting my money paying for GLM for FOUR MONTHS. It has NEVER been usable at all. Every single request gets the error "User Provided API Key Rate Limit Exceeded". This is so ridiculous. I can't even make a SINGLE REQUEST before getting rate limited. I thought it was a temporary bug so I continued paying for this garbage and it's still not fixed. STILL TOTALLY UNUSABLE.

I just tried again today by re-adding my API key in Cursor. I didn't make a SINGLE REQUEST FOR MONTHS. And as soon as I say "hi" to the model I'm getting rate limited.

Contacted support 2 times with screenshots and documentation. ZERO REPLY. They literally don't care, they take your money and run.

As others have pointed out, the 3X Claude usage is a scam.

"3× Claude's quota": a mathematical lie
Z.ai sells their plan promising "3× more quota than Claude Pro." What they carefully hide on their sales page: GLM-5.1 consumes quota at 3× speed during peak hours and 2× during off-peak hours. The 1× multiplier advertised as "permanent" simply does not exist.
Real result: at equivalent price, you get less usable quota than Claude Pro — while being limited to 1 single non-parallelizable session on GLM-5.1.
Opaque token counter
The dashboard displays token figures that are physically inconsistent with what the model can actually process. No way to know what is genuinely being consumed. Users are reporting "Token System Broken" on their own subreddit.
Email sent, ticket opened → complete silence. Their Discord is chaotic with no dedicated staff. For a company charging a monthly subscription, this is unacceptable.
Z.ai is an academic research lab that commercialized its models without building the customer service infrastructure to match. If your use is professional, look elsewhere. Anthropic or OpenAI APIs cost the same in real usage — with the transparency and support that come with it.
If you feel misled, open a dispute with your bank or payment provider for "service not as described" — it is your right.

Okay service for a good price.
I'm using their GLM Coding Pro plan and I'm not surprised that their models and the service are not as good as Claude for example. They're okay and I have happily used them for their value proposition, you get a lot of tokens to consume and while I don't rely on their models for complicated tasks they're very good work horse models.

Their latest models is not usable because of the slow performance during peak time

Using their coding plan, created a few personal projects and online apps. Yes, I do experience slow performance, especially with glm-5.1, but it accomplishes what I need it to. I'm looking for better, more stable performance in the future.

Well, I need to accept that I got scammed by z.ai

Purchased yearly Max Plan for over 600 USD and now they just nerfed the models. The GLM-5 is not usable if hosted by Z.ai. Any other provider serving GLM-5 delivers very good quality, but Z.ai serves a quanitzed model that delivers gibberish starting with 30k tokens.

Any mails sent to user feedback email got undelivered - they even dont care to ost a mail server.

Z.AI isn't delivering and banning people that are complaining on discord including myself. Ignoring the fact that they are not doing the job or accepting the fact that they are not giving users the promises they made!

GLM 5 is a excellent model. The only downside is the human factor. If you use the platform for some serious work and happen to identify an issue and call it out, you can kiss your hard work goodbye. I lost a couple of days of work on that platform - my chat history was wiped out clean by developers.

I have never been this disappointed with a purchase before.

Z.ai advertises that the Light plan offers 3x the usage of Claude Pro, and that the Pro plan offers 5x the Light plan — which effectively suggests you are getting 15x the usage of Claude Pro.

I purchased the Z.ai Pro plan today, and in less than two hours I had already consumed more of my five-hour quota than I typically use during four hours on Claude Pro.

It is completely fair if GLM-5 consumes more resources. That is not the issue. The problem is advertising that users get 15x more value, when in reality the experience does not even come close to matching Claude’s actual usable time.

In addition, the coding quality has produced significantly more errors compared to Claude Code in my testing. That makes the value proposition even harder to justify.

What frustrates me most is not the performance itself — it is the misleading marketing. If the plans had been presented transparently, I could have made an informed decision. Instead, I feel like I fell directly into an exaggerated marketing campaign.

Based on my experience, I cannot recommend Z.ai at this time.

Runs good for the first tries but after a while becomes really slow... Don't buy thinking it will be faster. It won't, it might even be slower actually!

I highly recommend Z.ai! I've used several AI platforms, but Z.ai is in a league of its own. The code agent generated complex scripts in minutes, saving me hours of work. The AI writer function consistently produces professional and readable content that exceeds my expectations. The platform is incredibly user-friendly, and the results are always top-tier. An exceptional tool for anyone needing fast, reliable AI assistance. The quality is consistently outstanding!

I have been using it for over a month and the quality of the service has been pleasantly surprising despite the low price

No recomiendo para nada si lo vas a usar con claude code en una terminal , la terminal no puede interpretar imagenes , page la version pro la cual si es verdad que tiene una duracion muchisima mas larga que claude , los tokens dura mas y lo he usado por 3 horas seguidas sin parar y no se acaba el limite , pero solo sirve para corregir errores sencillos , analizar contextos y nada mas , si tu le subes una imagen no la entiende y no la analiza , por lo cual esos 5 dolares de difencia entre Claude Pro y GLM PRO es mejor pagar claude.

The first session of the model will be amazing. I suspect that they use the original sonnet until the user has paid for several months or even a year. And then the model suddenly turns into something that is even worse than free alternatives. It literally does not understand basic concepts, html, and css become insurmountable. You run to support, but in response. No, we don't refund money. And it becomes clear why. It looks like fraud.