LLMBase Avis 

The inference API is too slow, for glm 5.2. They claim a tps of 100+ but the output is snail's pace, far far from 100 tps. More like 10tps.

Edit in response to the reply: Your models (/models) list advertises GLM-5.2 with "191 toks/s" in the model card.

As for the average provider claim, I disagree. TensorX, ollama, NeuralWatt all provide 100+ t/s.

I had a small issue with my account setup. Reach out to support and replied quickly. Was fixed within 30 mins and was my mistake to begin with. Appreciate they let me know what happened and it wasn’t an AI or copy-paste response.

The workspace functionality has been particularly useful for our team. Uploading documents and letting the AI reference them has saved us a lot of time when answering internal questions. The quality of responses improves significantly when company knowledge is available. We are excited to see how the feature evolves further.

Really good service where they are quick to reply if you have any questions.
Can highly recommend!

I switched over from ChatGPT and really like the variety of AI Models. And it's great that the subscription can be used with OpenClaw.

10000000% Scam!!! Stay away! You pay a subscription and dont get what you pay for. You pay for Deepseek4 pro and get Claude (likely Haiku VERY BASIC model). You get Chat GPT access equivalent to the free access you get without an account ! stay the hell away this is a total SCAM ! A total Grift! Thank GOD I did not give them any card details!

In reply to the below, I paid for a subscription without directly providing card details, that subscription got me a service was down/non functional for 90% of the time, with no response at all from the customer service/help. Several weeks later/ a month later I get a reply from somebody unapologetic as to the service I received. The idea is great but the service delivery is abysimal.