Chatllm Abacus Avis 11

I wanted to use this AI, but when I saw they were asking me to pay first without even offering a free trial or anything, I immediately checked the reviews. Why not, when you’re paying for the service before even using it? I’m glad I did that. I’m not buying it now.

I subscribed to ChatLLM by Abacus AI expecting access to a stable multi-model AI platform, but my experience was extremely disappointing.

First, the credit system feels completely unreasonable. With very light usage, I lost around 3,000 credits out of 20,000 in less than one day. It honestly felt like credits were draining in the background far faster than what the platform’s usage monitor was showing.

What made it worse is that the platform displays a small “credit usage” number after each response, but from what I observed, the actual balance reduction was significantly higher. That creates a serious transparency problem because users cannot clearly understand where their credits are going.

The second major issue was model availability. When I subscribed in the morning, the platform showed 33 models available. By the same night, it had dropped to 28 models — literally within my first day as a paying customer. Losing models only hours after subscribing is unacceptable and gives the impression of an unstable service.

I cancelled my subscription in less than 24 hours because the value simply was not there for me. Instead of focusing on using the platform productively, I found myself constantly watching credits disappear and questioning the accuracy of the usage tracking.

In my opinion, the service needs:
- much better transparency about real credit consumption,
- stable model availability,
- and a pricing structure that users can actually trust and predict.

Very disappointing experience overall.

Para mi es: Un Timo mas de los internets. una pena. pague 7 euros que no me devuelven y a las 4 horas de abrir la cuenta y crear un agente, ya me pedian mas dinero y quedo todo colgado. otro timo mas.

The app is not good and the credits system. Like only about 12 videos in A month. 33 credits if I say hi to Ai?. Not worth it.

I'm building an app and I'm just about finished, but ran into an issue. It's something "support" could easily do. They couldn't be bothered. They said: "just start over from scratch". Right. More time, thousands of credits, another 14 days of testing etc. Then, when I explained this, crickets. They ghost you. Pathetic and worthless. AVOID!

Abacus.ai ist zwar eine innovative und preiswerte Plattform und die eingegebenen Wünsche werden verarbeitet. Leider ist wie so oft das Ergebnis nicht immer zufriedenstellend, es gibt Abbrüche und die KI ist nicht immer schlauer als man selbst. Ich vergebe nur 2 Sterne, da ich ein komplettes Video mehrmals modifizieren musste, ohne den Erfolg wirklich zu sehen. Wenn es einen gab, hatte das Video einen kompletten Lila-Stich und konnte nicht verwendet werden. Dennoch wurden ohne Ergebnisse gleich einmal 23000 Credits fällig, die völlig umsonst waren. Die KI antwortete mir darauf, dass eine Gutschrift der Credits nicht erfolgen kann, da man ja dennoch generiert hat. So etwas finde ich sehr bedenklich. Trotz dem das Programm für das, was es anbietet, sehr preiswert ist, arbeitet man auf eigenes Risiko.

I want to warn potential customers about Abacus.AI and specifically the ChatLLM Teams subscription.

My experience has been extremely disappointing.

I explicitly asked for correct, step-by-step technical guidance with prior online verification. Despite multiple confirmations, I repeatedly received incorrect or not-verified instructions (including issues with Solana CLI installation and devnet faucet usage). This caused repeated errors, lost hours of work, and a completely unusable workflow.

But the biggest problem is the total absence of customer support.

Since November 5, I have contacted their support team several times — and received zero responses. Not a single reply. No follow-up. No assistance. Nothing.

I eventually filed a formal complaint and asked for a refund, but even that was ignored. For a product that claims to offer professional AI support, this level of customer service is unacceptable.

My advice:

Do not fall for this. The platform does not deliver what it promises, the quality is poor, and there is simply no support. You only lose money while plenty of alternatives exist that actually respond and provide real assistance.

At this point, I cannot recommend Abacus.AI at all.

They stopped the image generator once I reached 75% of my credit. This makes no sense since I still have 25% credit left!

In my experience, Abacus.ai was extremely disappointing. Immediately after creating an account, I was prompted to pay $10. Within two minutes of use, I was asked to pay another $10 or start a new session. After paying a total of $20, the system reported it was overloaded and halted my project. Later that same day, I tried again; the product quality—especially DeepAgent—was poor. I canceled the Pro plan that day, but the platform would not let me cancel the initial $10 upgrade and indicated it would auto-renew next month. I have screenshots, including card statements showing both charges, and can share them if needed.

This business feels like a scam. They claim they might throttle your usage if you use too many attachments, but I didn’t use any attachments at all. After just a couple of hours of simple text queries, 75% of my monthly credits were already gone. When I tried to generate my very first .png file, they immediately asked for more money. This kind of predatory behavior makes even the NZXT PC lending scam look tame. Completely unacceptable. Avoid at all costs.

De mi experiencia puedo sacar dos conclusiones. La primera, no usan los modelos reales, y la segunda, no mantiene el contexto.

Primero, lo que sorprende, es que nada más entrar a la web, e iniciar sesión, lo primero que te pide, es tu tarjeta. No te permite probarlo con un par de queries, ni siquiera observar la interfaz. De primeras, a pagar!

Lo segundo que más sorprende, es que las respuestas van bastante rápidas, demasiado rápidas incluso. En el caso del modelo 4o de chat gpt, va incluso más rápido que el modelo original, desde el propio chatgpt. Esto precisamente, me hace sospechar, que en realidad no están usando el modelo real, ya que incluso desde la API, la cual yo uso en mi trabajo, no va tan rápido. Esto sobre todo se nota en consultas complejas.

En cuanto a las respuestas, la primera pregunta que le haces, parece que funciona correctamente, responde más o menos a lo que le has pedido, aunque a veces se pierde. Es cuando haces las segunda, tercera, y más preguntas cuando te das cuenta de una cosa: NO USA CONTEXTO.

El contexto es lo principal en un chatbot, y en mi experiencia, ChatLLM, parece que cada prompt es independiente de los anteriores. Si le preguntas, sí que parece que se acuerda de lo que le dijiste, pero a la hora de la verdad, si tu no le está escribiendo en cada prompt, que se acuerde de seguir las normas que ya le has especificado anteriormente.

Ejemplo de mi experiencia própia:
Yo le pido que me traduzca de Alemán a Español la canción X, y que me responda con los versos línea por línea en los dos idiomas (estoy aprendiendo alemán)

La primera vez, lo hace más o menos bien. El problema es cuando mi siguiente mensaje es "Ahora Rotkäppchen" (que significa caperucita roja en alemán).
A lo que el ChatLLM me responde, contandome de qué va el cuento de caperucita roja.

Este es solo un ejemplo, pero esto me pasa desde el día 1 que lo uso, lo he usado 2 días, pero no puedo más, si no mantiene el contexto, el chatbot es inusable.