A request is any message you send to the chatbot. Each month you have a fixed package of requests available. Within it, you can communicate with AI. For example, simple AI models spend one request from your package per your message. More advanced neural networks may use, for example, three requests for one message. This is because they: Are more powerful and smarter Require more computational resources to generate a quality response Have higher cost price and final price

What to do if requests run out?

There are two options: If you regularly hit the limit and understand it won't be enough going forward, we recommend switching to a higher plan If you need more requests only occasionally, you can purchase them separately

What neural networks do you have?

We add almost all models that can be useful in everyday life for most users. We monitor new releases and add them promptly — usually within one or two days. We also keep old versions of neural networks for a while in case the new one proves unstable. The full and up-to-date list of available neural networks can be viewed here

How to cancel the subscription?

You can cancel your subscription at any time: In the bot: use the /unsubscribe command and follow the instructions In the web version: click on your profile icon in the top right corner, go to settings, and find the “Unsubscribe” button But it's better not to cancel anything, of course, because soon robots will take over the world, and we don't see much point in resisting 😉

Do you have technical support?

Of course, it can be reached either by email at support@chataibot.ru or @ChatAISupport

Can multiple people be added to the subscription?

Yes, this option is available in the “Business” plan. Up to 30 users can be added to it. Despite the name, any individual can subscribe to this plan. It's perfect for a family or small team, for example.

Who has access to my chats, photos, and videos?

No one except you, us, and the neural networks you use. We don't grant access to anyone else; your privacy is under your control.

A Guide to Neural Networks 2026: Which Digital Assistant Should You Choose?

Today’s AI world is no longer just chatbots, but fully-fledged digital personalities. Each has its own advantages: one is good at coding, another is a good creative writer, and a third is an excellent psychotherapist. We’ll help you choose and tell you more about the most relevant models for 2026.

Universal — for any task

1. GPT-5.4 — “Master of Code and Tables”

If your work involves programming, complex automation, or heavy Excel spreadsheets, this model is your best friend. It has phenomenal long memory and maintains a leading position in professional developer benchmarks.

Stats: Leader in programming tests (59% in SciCode) and technical command execution (54% in Terminal-Bench).

Real-Life Example: You have a family budget in Excel that has devolved into chaos. You upload the model file, and it automatically corrects formula errors, creates beautiful expense graphs, and writes a script that automatically pulls data from your bank statements.

2. Claude Sonnet 4.6: “The Diligent A-Student”

If you give clear instructions like “do A, then B,” Sonnet will execute them more accurately than anyone else. It doesn’t overthink things and strictly adheres to your boundaries and rules.

Stats: One of the best results in following complex rules (79% in IFBench) and high intelligence at high speed (63 points in Agentic Index).

Real-life example: You have 50 recipes, all written down “on the fly.” You ask Sonnet to organize them into a single format: first ingredients in grams, then calories, then cooking steps. It will do it perfectly, without missing a single comma.

3. Gemini 3.0 Flash — “Jet Assistant”

A sports car among neural networks. It won’t take long to figure things out, but will deliver instant results. Ideal for small tasks when time is money.

Statistics: High accuracy (54%) among all fast search-optimized models.

Real-life example: You’re in a store and see a TV on sale. You ask the model, “What are the reviews for this model over the past month, and where is the cheapest place in town to buy one?” You receive a response with links in a second.

Another example: You need to quickly summarize a long article or check an email for errors while rushing to a meeting. You enter the entire text into the chat and receive a completed report within seconds.

Advanced – for serious tasks

1. Claude Opus 4.6: “The Wise Writer”

The most “human” and creative model on the market. She doesn’t have robotic pathos—she writes beautifully, deeply, and understands your true meaning better than anyone else. She also has the highest “agent” potential. This means she understands better than anyone how to execute a complex chain of actions without your supervision.

Stats: #1 in the world in the Agentic Index (score 68). She understands how things work in reality better than anyone else (GDPval — 57%).

Real-life example: You need to organize a large family celebration for 30 people. You give Opus the guest list and budget. She creates the menu herself, calculates the ingredients, finds delivery links, and writes polite invitations to each guest, taking their food preferences into account.

Another example: You need to write a script for a video or a heartfelt greeting that doesn’t look like an online template.

2. GPT-5.4 High — “The Master Thinker”

A benchmark for logic and deep analysis. It can handle huge volumes of information without losing the thread.

Stats: Leader in long-form context retention (74% in AA-LCR) and scientific knowledge (92% in GPQA Diamond).

Real-life example: You’re presented with a complex medical report or a 40-page apartment purchase agreement. The model will analyze the document, highlight all hidden risks, and explain complex terms in simple terms so you don’t make mistakes.

3. Gemini 3.1 Pro — “The Walking Encyclopedia”

The world’s smartest model in terms of knowledge and visual perception. If you need to find something or understand what’s in a picture, this is the one for you.

Stats: The absolute leader in image recognition (82% in the MMMU Pro test) and erudition (55% accuracy in world knowledge). It manages third-party services almost perfectly (98% in T2-Bench).

Real-life example: If you take a photo of an old car part, Gemini will immediately tell you what it is, how much it costs, and where to find repair instructions.

Google’s neural network is the best at Googling something or verifying information. The portal artificialanalysis.ai ranks Gemini first in the world for intelligence. Not sure which model to choose? Go for it.

Quick – when you need to find something or come up with something urgently

1. Perplexity – “Digital Librarian”

This is the search engine of the future. Instead of a list of links, it provides a ready-made answer, supported by evidence.

Advantages: Each statement is accompanied by a link to the source.
Real-life example: You want to find out whether vitamin D really helps with colds. Perplexity will collect data from recent medical studies and provide a short answer with links to each article.

Other models can also provide links to sources—but only upon request. They are also more likely to make errors.

2. Claude Haiku 4.5 — “The Honest Creative”

A small model that almost never lies. Ideal for simple creative tasks.

Stats: #1 in the world for truthfulness (74% in Non-Hallucination Rate). It’s the least likely to invent non-existent facts.
Real-World Example: You need to come up with 20 funny greetings for a colleague or quickly translate a short email into a foreign language. Haiku will do it instantly and in a very “live” language.

3. GPT-4.1 nano — “The Free Workhorse”

Your insurance in case you run out of powerful models. Our Chat AI portal offers unlimited use of the model.

Advantages: Reliable in basic tasks: correcting typos, reformulating text, or compiling a shopping list.

Real-life example: You’ve written a letter to your child’s teacher or a complaint to the housing office. Nano will quickly correct your style, remove errors, and make your text more persuasive.

Results: These are just 9 of 39 text models presented on our website. To understand which neural network is right for you, you should try working with at least two of them. By the way, on our website you can simultaneously send the same request to several text models and compare which answer you like best (and therefore, which neural network is more suitable for you). Give it a try!