Text Neural Networks Guide: How to Use ChatGPT, DeepSeek, and Other Chatbots

Just a couple of years ago, ChatGPT was practically the only powerful text neural network available to a wide range of users. But today everything has changed: dozens of alternatives have appeared, each with its own features, strengths and weaknesses.

We collected information about popular chat bots: how to create an account, what functions they have and how much a subscription costs. We also tested neural networks on three identical prompts that demonstrate the ability to logically, creatively and explain complex concepts.

Key terms

LLM: short for Large Language Model. This is the correct name for neural networks created to generate text.

Multimodality: the ability of a neural network to perceive images, voice and other types of data in addition to regular text. For example, the OpenAI GPT-4o neural network is able to understand what is depicted in an uploaded image and also support a dialogue in audio format. >Reasoning mode: the ability of the neural network to solve a query step by step, spending more time and computing power on forming an answer. This helps to get a more accurate result in logical and mathematical problems.

ChatGPT

Applications: website, iOS, Android, Windows, macOS

This neural network started the chatbot boom. The developer of ChatGPT, the OpenAI startup, regularly rolls out new features. For example, an even more capable generative model GPT-4.5 has recently appeared.

Some features are available only by subscription. For regular users, there are two plans: Plus for $20 per month and Pro for $200 per month. OpenAI also provides plans for companies and educational organizations. And the models are also available to developers via API.

You can use the basic functions of ChatGPT even without an account. But to get all the features, you will have to create an account on the website or in the application. You will need an email address and a phone number that belongs to a country in the ChatGPT coverage area. Just don't forget to set up two-step authentication after completing the registration so that no one else can access your profile.

ChatGPT's main functions and modes:

• Continuous voice mode, almost indistinguishable from a live conversation.
• Image generation (via the DALL-E illustrative model).
• Image and document analysis.
• Internet access for up-to-date information.
• GPTs: customized versions of ChatGPT, tailored to perform specific tasks.
• Deep Research mode, in which the service conducts a detailed study on a given topic using data from the Internet. Available only on paid plans.
• Operator mode: ChatGPT opens a browser and performs actions on third-party sites for you. Available only on paid plans.
• "Reasoning" ability.
• Code execution: ChatGPT can show a working prototype based on the generated code or write and execute a Python program to provide an exact solution to a math problem.
• Built-in Canvas editor for working with neural network responses without leaving the ChatGPT interface.

In general, OpenAI has released so many models that it is already difficult to understand them. Here is a hint table for the company's text neural networks:

But if you are not going to dive into the nuances of ChatGPT, it is enough to know the following:

1. By default, the free plan uses the GPT-4o model. It supports multimodality.
2. When the limit for using GPT-4o is exhausted, you will be switched to a simpler neural network 4o-mini.
3. The "Reason" function activates the o3-mini model. It is capable of "reasoning", but does not support multimodality.

DeepSeek

Applications: website, iOS, Android

At the beginning of 2025, the Chinese startup DeepSeek made a splash: its eponymous chatbot copes with most tasks at the level of GPT-4o from OpenAI, while being much cheaper to train and maintain.

DeepSeek attracts with the absence of paid features. The company makes money by providing developers with access to the API. But due to high demand, the neural network regularly refuses to respond, citing server congestion.

To create an account, you only need an email address. You can also sign in through a Google account, and in the iPhone app, also through an Apple Account.

Main functions of DeepSeek:

1. The ability to "reason" (the "Deep Thinking" function).
2. Access to the Internet to obtain up-to-date information.
3. Analysis of text documents. You can also upload an image, but the neural network can only recognize the text in it.

The DeepSeek-V3 generative model and the DeepSeek-R1 reasoning model are open source, meaning you can run them on your own hardware without having to worry about privacy.

Claude

Applications: website, iOS, Android, Windows macOS

Another strong player in the neural network market is the Anthropic startup and its LLM family called Claude. The most advanced neural network at the moment is Claude 3.7 Sonnet.

You will need an email address and a phone number to register. You can also log in via a Google account. But right now Claude warns that it is temporarily not registering new users.

Claude is especially strong in code generation and creative tasks. There is a choice of response style right in the interface. Like ChatGPT, Anthropic can execute code to solve problems and show a preview of the project based on the generated code. You can even connect a GitHub repository so as not to copy the code into the neural network manually. But Claude cannot go online for fresh information.

Other Claude capabilities:

1. Image and document analysis.
2. "Reasoning" mode (only in the paid version).

A Claude Pro subscription costs 18 euros per month. With the subscription, the "thinking" mode is activated, the limits on the number of requests are increased, and experimental functions appear. You can also connect to Claude via the API interface.

Grok

Applications: website, iOS, Android

It's hard to name a field in IT where Elon Musk hasn't made his mark. His startup xAI is engaged in research in the field of artificial intelligence. The third version of the main neural network xAI, Grok, is currently relevant.

Grok's main functions:

1. Internet access.
2. DeepSearch mode for conducting research on a given topic using data from the Internet.
3. "Reasoning" mode.
4. Image generation.
5. Document and image analysis.
6. Advanced Voice Mode (for paid subscribers).

Grok's pricing structure is a bit confusing. The highest limits are available with the $40/month X Premium+ subscription, but there's also a $30 SuperGrok plan that doesn't give any benefits to your X account. You can also connect to Grok via API.

Gemini

Applications: website, iOS, Android

Google also has its own AI assistant. On Android smartphones, it is gradually replacing the "Google Assistant", but there are interfaces for both iOS and the web.

Google is more scrupulous than others about users from countries where Gemini is not officially available. To access the website or the chatbot application, simply changing the IP address is often not enough. There is no guaranteed way to get into Gemini: you will have to experiment with Google accounts, browser or smartphone settings.

But if everything works out, then the following functions await you:

1. Internet access.
2. Image and file analysis.
3. Deep Research — in-depth research on a given topic. Available on a paid plan.
4. Advanced voice mode.
5. "Reasoning" mode.
6. Image generation using the Imagen 3 model.

Gemini can also connect to third-party services such as Google Workspace or Spotify. For example, you can upload a document stored on Google Drive to the model.

The interface has a switch between models. Today, the most powerful neural network that is open to free users is Gemini 2.0 Flash. There are also its "reasoning" subtypes that work in experimental mode.

Google has a Gemini Advanced subscription, which opens access to advanced models, and also greatly increases usage limits. For example, with a subscription, the context window of the neural network increases to a million tokens - several times more than competitors. In practice, this means that Gemini is able to "digest" a larger amount of information at a time. For example, you can upload a very large document or book for analysis. The subscription costs $ 19.99 per month.