ChatGPT overview: models, features, prices explained

ChatGPT started the generative AI revolution. Since then, it has grown tremendously. Today, it’s the number one AI platform and offers a multitude of tools, services, and plans. Moreover, because the industry is in such a dynamic phase right now, things keep changing.

In other words: It’s hard to keep up with the latest developments, to understand what is possible, and to know how to use it.

If you are a creative working in a field like Marketing, Online Marketing or Content Marketing, this page is for you.

With it I aim to give you a comprehensive overview of ChatGPT and its many features and offerings.

General overview

ChatGPT is a product of OpenAI. It is their most successful and influential offering to date. OpenAI started out as a pretty obscure research lab that wanted to develop AI to the benefit of humanity without worrying about revenue and profit.

When ChatGPT became an instant hit in late 2022, this changed dramatically.

As you will see below, ChatGPT is not just one product. It does several things, it does them in different ways, and it offers many ways to use it.

There is a free version with limitations. There are several paid levels as well as special offerings for big companies and education. Last but not least, there are options for developers that can be relevant for non-developers as well.

The core of ChatGPT is text. But it can also analyze and generate images. You can talk to it and it will respond in a human-sounding voice. It can do in-depth research for you.

That doesn’t mean that it’s always the best choice. There are good alternatives. Claude is generally better for writing and is an excellent assistant. Midjourney can be a better choice for creative image generation. Googles Gemini is the number one when it comes to multifunctional AI platforms.

But still: ChatGPT is the household name. OpenAI is keeping the pace of innovation high, especially since Google’s swift progress has put them on high alert.

If you want to read more: I published another in-depth overview comparing ChatGPT, Gemini, Copilot and Claude to find the best AI für content creation.

With all of that out of the way, let’s have a closer look at everything ChatGPT offers.

ChatGPT’s models and options explained

When you talk to ChatGPT, different models (versions) are available. And you can also choose how quickly or how intelligently you need the AI to reply. Yes, there’s a trade-off: You can have instant answers or smart answers. There’s also a middle option for the undecided.

This is what that looks like in the “model picker” menu attached to the right side of the prompt bar:

Behind the scenes, two things happen: For one, slightly different models get to work depending on your choice. Secondly, these models might take a moment to consider your request first and map out how to reply (often called “reasoning” or “thinking”) or they answer right away.

If you’re using ChatGPT for creative tasks and writing, there’s another consideration: The “Instant” option often produces more interesting and creative results. The “High” option is better suited for strategic or conceptual tasks.

A general example: When asking for five different ideas for a headline, I would choose “Instant”. If I wanted ChatGPT to take my chaotic list of random ideas and form them into the first draft of an editorial plan, I would choose “High”. Some more examples to help you choose:

• Instant: writing and editing, summarization, translation, everyday Q&A, brainstorming
• Medium: planning, analyzing documents, comparing alternatives
• High: research synthesis, difficult planning tasks, problems where accuracy matters more than speed

Fair warning: This menu has changed many, many times over the last year. The people at OpenAI never seem to be fully happy with their latest version. I’m trying to keep the screenshot and my explanations up-to-date, but it might already look slightly different (again). But the general gist of the choices hasn’t changed much, just their presentation and naming.

There’s one additional option that used to be in this menu but has since moved into Settings → General: “higher intelligence”. With this activated, ChatGPT can switch to a higher level of thinking if the request seems to require it.

I would keep this activated and let ChatGPT decide for you. But if you prefer to make this choice yourself, maybe because you want “Instant” to always be instant, then that’s the setting you are looking for.

Additionally, a quick note on version churn: the model actually doing the work behind the scenes changes frequently. It’s moved through GPT-5.3 and GPT-5.4 since the beginning of 2026, and as of late June 2026 GPT-5.5 Instant is the default, with OpenAI reporting meaningfully fewer hallucinated answers on fact-heavy questions like medicine, law, and finance. As you can see above: These older versions are still available if you want them, even the “ancient” o3. If you don’t have a specific use case for these models or don’t even know what the differences might be, you can safely ignore them. OpenAI tries to accommodate users who have gotten accustomed to a specific model instead of discontinuing them right away.

And it doesn’t end there: OpenAI has, after a brief preview period, announced the next version, fittingly called GPT-5.6, in three variants it named Sol, Terra and Luna. GPT-5.6 Sol is now available for the “Medium” and “High” options.

If you want to know about these and other current developments and always get the lates tips and tricks, subscribe to the bi-weekly Smart Content Report newsletter!

Search and research with ChatGPT

One of the interesting use cases of ChatGPT is to research a topic, an idea or a question.

In the beginning, ChatGPT could only access its “world knowledge”: the information the AI has encountered during its training process.

Since this training process takes many months, this knowledge is often outdated. Even worse: it might have been incorrect from the start, because the training material was not fact-checked.

Additionally, there’s the phenomenon of “hallucinations”: This term describes instances when an AI adds information that fits well into its reply, but is made up. I show you in another article how to spot and prevent AI hallucinations. I also explain what causes them.

This combination of problems made ChatGPT and others deeply flawed tools for research.

That has changed thanks to three innovations:

1. Web access

Today, you can ask ChatGPT to look something up online or give it a URL to read. At times, it might also decide on its own that it needs to look for current information online.

It will give its sources at the end of the reply. It is a good habit to check these, because an AI can misunderstand or misrepresent information or still “hallucinate” it. It is overall less likely. But the chance is not zero.

2. ChatGPT search

In addition to that, OpenAI added ChatGPT search. This is a separate feature you activate by choosing the “Web search” option in the “+” menu on the lefthand site of the prompt bar. It has a small globe icon.

ChatGPT will then search the web for information regarding your request and present it on a results page with a summary, as well as images and videos where useful.

It lists its sources as well.

ChatGPT search is helpful for getting a quick first overview of a topic.

OpenAI modeled this feature after the AI search engine Perplexity. Another example of the same idea is Google’s “AI mode”.

3. ChatGPT deep research

The third option is ChatGPT’s “deep research” feature. This is an automated (“agentic”) research assistant. It scours the web for information in relation to your request and presents its results in a comprehensive report that includes images and links in addition to text.

This can be an immensely helpful tool for finding facts, figures, quotes, and more on a given topic.

A deep research session can take several minutes, even up to half an hour. You don’t have to wait for the results. You’ll get a notification when it’s done.

Since early 2026, deep research also lets you point it at specific websites instead of the whole web. This is useful if you only trust a handful of sources for a given topic. You can now also edit its research plan and redirect it mid-task instead of waiting for the final report to see it went the wrong direction. OpenAI’s help center has the details.

In case you want to know more: I’ve written a comprehensive article on how to use ChatGPT Deep Research in Marketing.

I’ve also published a comparison of Deep Research tools. ChatGPT’s is actually not my favorite and not the one I keep using. Find out in the comparison which one’s my go-to Deep Research service!

Generate images with ChatGPT

For the longest time, ChatGPT’s image generator was called “Dall-E” and a separate AI. In fact, Dall-E predates ChatGPT.

In March of 2025 OpenAI integrated image generation directly into its GPT-4o model. Its official name was GPT Image 1. The next version arrived in December 2025 as GPT Image 1.5. It sped up what had until then been a fairly sluggish image generator. The latest version is now GPT Image 2, also known as ChatGPT Images 2.0. It expanded the feature’s capabilities once again.

Without getting into the technical details: It works fundamentally different from several other image generators like Midjourney, Stable Diffusion, Flux or Ideogram.

I explain in a separate article what makes ChatGPT’s image generation so special. I also show its capabilities with several examples. In my article on the latest version, you’ll find an infographic that demonstrates two capabilities: ChatGPT can now plan and execute more complex visualizations on its own, and research the numbers and facts behind them beforehand. That’s especially exciting for content creation!

Because of its unique approach, it has unique strengths and weaknesses:

Strengths:

• Can handle complex image ideas with several different elements.
• Much better at rendering text correctly, even a lot of text.
• Good command of image styles and switching from one to the other. This has lead to the image generator going viral.
• Easy to use, because you don’t need to learn any special syntax.

Weaknesses:

• No advanced features and options to fine-tune results.
• Not as good at photorealism as other options.
• Still gets details wrong, doesn’t understand prompts or gets confused in longer chats.
• Can be slow.

In my experience, ChatGPT’s image generation is an especially useful tool for the needs of marketers. One reason is its capability to display lots of text correctly. Another one is how it can handle visuals that are more graphics-heavy and complex.

At the same time, Googles image generator (aka “Nano Banana”) has caught up to ChatGPT. It’s available in the Gemini platform.

Other alternatives also still have their place when it comes to photorealism or more artistic styles and creative expression.

Read my in-depth coverage of the AI image generation market. It features several interesting offerings, explains their differences and use cases, and also talks about the important topic of legal issues with these tools.

Features that make ChatGPT more useful

Over time, OpenAI has added several useful features that make working with ChatGPT easier.

Writing blocks (Canvas successor)

Problem: When working on a text, the draft and the discussion about it are intertwined. This makes it confusing and awkward to work on a piece.

Solution: The solution for this used to be Canvas. It was a special view that showed the chat on the left and the document on the right.

It was one of my favorite features and also ChatGPT’s best kept secret for writers. It was very useful, but hard to find. Instead of improving it or making it more easily accessible, OpenAI got rid of it all together. The current GPT-5.5 models don’t have it anymore. You can still use it with older models as long as they are offered.

The successor is called “writing blocks”. Besides the awkward naming (writer’s block anyone?), it’s not a true replacement.

“Writing blocks” appear directly inside a chat response, similar to an embedded document. When ChatGPT produces something you’re likely to revise, e.g. an email, a blog draft, a script, it now places that content in this editable block right there in the conversation.

How to use it: You can edit the block directly, ask ChatGPT to revise it, and for code blocks, preview or run supported languages (HTML, React, SVG, Mermaid diagrams, Python) without leaving the chat.

Tip: Version history and true side-by-side comparison, some of Canvas’s best features for serious editing work, aren’t part of writing blocks yet. If your workflow relied on those, this is a downgrade for now, and you’re not alone in thinking so: it’s been a common complaint among developers on OpenAI’s own community forum since the change.

OpenAI introduced Canvas in October 2024 and replaced it with writing and code blocks starting late May 2026.

Saved memories & chat history

Problem: Every new chat with ChatGPT starts at zero. It doesn’t know anything about you or prior discussions.

Solution: First, OpenAI added a way for ChatGPT to save information it deems important as “reference saved memories”. The AI can access this across chats. In the next step, it made all other chats available to ChatGPT with “reference chat history”. In June of 2026, OpenAI rolled out a significant memory upgrade internally called “Dreaming.” ChatGPT now runs a background process that reads across your past chats and keeps a synthesized, continuously updated picture of you, including revising outdated facts automatically. You can review and correct what it has inferred on a dedicated memory summary page. The rollout started with Plus and Pro users in the US, with Free, Go, and other countries following in the weeks after launch.

How to use it: You can ask ChatGPT to add a fact to its memory. You can access all saved items and delete them in the settings. In addition to this, access to prior chats happens automatically, but can also be triggered by explicitly asking ChatGPT to take into account everything discussed so far. All of this can be enabled or disabled in settings.

Tip: Especially the access to prior chats can open up new possibilities. For example, ask ChatGPT about your personal strengths and blindspots while taking into account everything it knows about you. Or, when starting a new project, it can remember certain facts about you, your background, your target audience, your preferences etc. Since memory is now synthesized automatically rather than only written by you, it’s worth glancing at the memory summary page every so often to correct anything ChatGPT inferred wrong.

Memory was added in February 2024. The ability to reference chat history was announced in April 2025. “Dreaming” was added in June 2026. Learn more about all of this in these FAQs.

Projects

Problem: When you need to start several separate chats to work on a bigger task, you need to add background information, instructions, and files every single time. It can also be difficult after a while to find these related but separate chats in your chat history.

Solution: Projects allow you to group chats into a folder and give you the option to add instructions and files that ChatGPT will automatically access for each new chat.

How to use it: Use the “+” when pointing at the “Projects” heading in the main menu. Name your project and add instructions and files.

Tip: There’s a maximum amount of files you can upload to a project and you’ll get an early warning when approaching it. Too many instructions and files might lower the quality of ChatGPT’s answers.

Projects were introduced in December 2024. Learn more about it in this official support post.

GPTs

Problem: When you have a specific, recurring task, you need to provide the same detailed instructions, background knowledge, and desired tone to ChatGPT every time. Friends, family, and colleagues might have the same need.

Solution: GPTs are custom versions of ChatGPT that you or others can create for specific purposes. This is why they are sometimes called “Custom GPTs”. Such a GPT can be pre-loaded with specific instructions, extra knowledge, and a combination of skills like web browsing or image creation, making it an expert for a particular task right from the start.

How to use it: Paid subscribers can create their own GPTs using the GPT Builder, which guides you through the process conversationally. No coding required! All users can find and use GPTs created by the community in the GPT Store. OpenAI has a FAQ about how to publish your GPT there. But you can also share your GPT just with specific people or groups in your workspace. Or you decide to keep it all to yourself and don’t publish it at all. That is also an option.

Tip: Before building your own GPT for a common task, browse the GPT Store. You might find ready-made tools for things like creating presentations with Canva, automating tasks with Zapier, or learning a new language.

GPTs and the GPT Store were introduced in June 2024.

ChatGPT Work

Problem: A regular ChatGPT chat is limited to the conversation itself. If you need something more substantial, e.g. a competitor analysis, a status report pulled from several tools, or a full slide deck, you still have to gather the source material yourself and stay at the keyboard while ChatGPT works through it step by step.

Solution: ChatGPT Work is a new agent mode that sits alongside the regular Chat and Codex modes. It can connect to the apps and files where your work already happens, stay on a complex task for hours by breaking it into smaller steps, and hand you back finished material such as a document, spreadsheet, presentation, or even an interactive “Site” you can share via a link. It runs on OpenAI’s GPT-5.6 model, and you can also set it up to run repeatedly or on a schedule.

How to use it: In the ChatGPT desktop app, switch between Chat, Work, and Codex from the menu in the top left. On mobile, use the dropdown at the top of the screen to switch from Chat to Work. Before you start, connect the plugins for the tools you rely on, e.g. Slack, Microsoft Teams, Google Drive, SharePoint, Gmail, or your calendar and CRM. ChatGPT pulls context from these automatically, or you can point it at a specific one by typing “@” followed by the app’s name. Describe your goal, review the step-by-step plan ChatGPT proposes, and approve it to start the work. You stay able to check progress, answer questions, and approve important actions along the way.

Tip: Work is built for longer, more involved projects, not for quick questions, so it draws on your plan’s usage differently than a normal chat. Pair it with Scheduled Tasks if you want it to check in on something on its own, e.g. turning new Slack updates into an updated status document every week.

ChatGPT Work was introduced on July 9, 2026, alongside GPT-5.6. It’s rolling out first to Pro, Enterprise, and Edu plans, with Plus and Business following a few days later. Learn more in this OpenAI Help Center article.

Custom Instructions

Problem: You have to repeat key information about yourself or your preferences (like your profession, your target audience, or a desired tone) in every new chat.

Solution: Custom Instructions lets you save standing information and guidelines that ChatGPT will automatically consider in every new conversation, so you don’t have to provide that context again.

How to use it: In your settings, find the “Personalization” or “Customize ChatGPT” section. There you can add your instructions and enable the feature for all new chats.

Tip: Use the two provided fields strategically. Put stable facts about yourself in the first box (e.g., “I am a content marketing manager”) and instructions for how ChatGPT should behave in the second (e.g., “Use a friendly, engaging tone.”).

Other notable features

• Scheduled Tasks: Ask ChatGPT to get back to you at a regular interval.
• Desktop: ChatGPT on your computer. An app for Windows and Macs that not only brings ChatGPT to your computer, but can also work and interact with content on your device. It unifies the chat, „Work“ and „Codex“ features and is also meant to integrate a successor to the Atlas browser.
• ChatGPT Translate: A dedicated translation tool. A standalone page, separate from the main chat, for quick translations across 50+ languages, with one-tap options to make the result more formal, more fluent, or simpler. No account required for basic use. Launched quietly in January 2026.
• Codex: OpenAI’s coding agent. Originally built for developers, it’s increasingly used well beyond them. This was probably one motivation behind the introduction of ChatGPT Work (see above). Learn more about AI agents and how they can assist you in this article.
• ChatGPT Record: Let ChatGPT transcribe and summarize audio recordings. An experimental feature only available in the Mac desktop app at the moment.
• GPT-Live (formerly Advanced Voice Mode): Talk to your AI assistant. An impressive achievement with limited relevance for everyday tasks of creatives.
• ChatGPT Atlas: A browser specifically made for ChatGPT. OpenAI announced in early July that they will discontinue Atlas in favor of a similar feature in the ChatGPT desktop app.

ChatGPT prices and access

OpenAI makes it challenging to fully understand what the different levels of access include exactly. The pricing page offered by clicking on “upgrade” is vague (and only available to logged-in users). But there’s another pricing page:

As you can see: It only includes relative language and no concrete numbers. This gives OpenAI the freedom to redefine at any moment what constitutes “limited” versus “expanded” or “maximum”.

To make it slightly easier to read, I collected this information in a table of my own:

OpenAI’s subscriptions and their features

AI Model or Feature	Free ($0/month)	Go ($8/month)	Plus ($20/month)	Pro (from $100/month)
Reasoning	Limited	More	Advanced (GPT-5.5 Thinking)	Pro (GPT-5.5 Pro)
File uploads	Limited	More	Expanded	Unlimited
Messaging	Limited	More	Expanded	Unlimited
Image generation	Limited and slower	More	Expanded and faster	Unlimited and faster
Memory and context	Limited	Longer	Expanded	Maximum
Deep Research	Limited	Limited	Expanded	Maximum
Agent mode	–	–	Expanded	Maximum
Projects	Available	Available	Available	Expanded
Tasks	–	Available	Available	Expanded
Custom GPTs	Use	Create and use	Create and use	Create and use
Sora video generation	–	–	Limited	Extended

ChatGPT Free

Not only can you use ChatGPT for free, you can also use it without an account. Unsurprisingly, this service is limited in several ways. Some features and options are not available at all, while others only give you a taste of what is possible. Features like image generation or access to advanced models is especially minimal.

Note: It makes a difference whether you are logged in or not. One advantage of having an account is that you get a history of all your chats. You also need to be logged in to use custom GPTs, for example.

ChatGPT Free gives you an idea of what is possible and may be enough for occasional use. Bear in mind that there will be a downgrade to a less capable model if you use it a lot. You will also encounter hard limits that will make you wait until you can use it again.

There’s a bit more information to be found in the official FAQs. According to that page, features included in free are:

• Search the web and obtain up-to-date information
• Analyze and extract insights from your data
• Upload images or files in your prompts
• Discover and use GPTs
• Create images in ChatGPT
• Use Library with 500 MB of storage

Just be aware that ads are now part of the deal: OpenAI began showing them to logged-in adult users on Free and Go in the US starting February 9, 2026. They show up as a labeled block below ChatGPT’s answer, are separate from the chat model itself, and you can dial back personalization in Settings → Ad Controls. If you don’t want ads, you need at least ChatGPT Plus.

ChatGPT Go

ChatGPT Go started as a trial in India, was soon expanded to more Asian countries and then taken global in January of 2026. With $8/month it is an affordable way to use ChatGPT more often.

If you look closely, ChatGPT Go has a very similar feature set to ChatGPT Free, but with much higher limits. In other words: Switching from Free to Go doesn’t give you much additional functionality, but you can use it more. Whereas if you switch to Plus you gain additional features.

According to OpenAIs support page ChatGPT Go offers:

• “Unlimited” chat access to GPT-5.5 Instant (subject to “abuse guardrails” and any model-specific usage allowances)
• Extended access to image generation
• Extended access to file uploads
• Extended access to advanced data analysis
• Longer memory for more personalized responses
• Access to projects, tasks, custom GPTs, and Library (up to 4 GB)

ChatGPT Go seems like a natural progression from the Free tier if you run into its usage limits regularly. It could also be worth a downgrade for current Plus users who don’t need any of the features only available there.

Just one heads-up: you might also encounter ads.

ChatGPT Plus

ChatGPT Plus is the next paid tier for individual users. This is the option you are most likely to consider if you use ChatGPT professionally. At the time of writing it costs $20/month. For this amount you get more access to more advanced models and generally higher limits. However, there are still limits in place.

One heads-up if you haven’t checked in on ChatGPT for a while: GPT-4o, the model many users had grown attached to for its warmer tone, was retired in February 2026 despite a vocal user backlash and petitions to keep it around.

According to the official help page you get:

• Priority access during high-traffic periods
• Access to higher GPT-5.5 limits
• Access to advanced reasoning models
• Faster response speeds
• Expanded features:
  • Voice conversations
  • Image generation
  • File uploads and analysis
  • Deep Research tools (where available)
  • Custom GPT creation and use

ChatGPT Plus is a good way to use this assistant and all its features for real work every day. ChatGPT Go could be an alternative for you if you are okay with ads.

ChatGPT Pro

ChatGPT Pro is by far the most expensive tier: $100 or $200/month. This is for heavy users, especially of the most advanced models.

According to OpenAIs official information, the difference between the two Pro options are usage limits. For $100/month you get 5x the usage included with ChatGPT Plus while for $200/month it’s 20x.

The big jump in price makes this only viable for people with complex tasks, who need the most advanced and powerful AI models on a regular basis, and who use it a lot per day.

ChatGPT Business

If you need more than one account, have a look at ChatGPT Business (formerly “Team”). It will cost you $20/month per member with annual billing or $25/month per member with monthly billing. As you would expect, you get an admin panel. The limits for some features are higher than ChatGPT Plus (no details given).

ChatGPT Enterprise

If you are in a large organisation, ChatGPT Enterprise may be the right choice for you. It includes more options and settings. As usual, no pricing information is available.

ChatGPT Edu

As the name ChatGPT Edu implies: This is meant for universities. No pricing information is available.

Not only for developers: API access

Last but not least: There’s a way to access OpenAI’s services, including its AI models, without using the official website. The keyword here is Application Programming Interface (API). This is meant for developers who want to build a product using OpenAI’s offerings. Learn more here.

But it’s also available to you. And you don’t have to be a programmer to find it useful.

The main advantage of API access is its pay-per-use pricing: Instead of paying a fixed amount per month, you pay for what you use. This can be much cheaper, especially for single users.

Also, there are not the same limits that you might find with the “Plus” account.

One downside: Special features like Deep Research are not available this way.

A simple way to use this is one of many plugins for WordPress that bring AI into its backend. Examples include AI Engine and AI Puffer. Or you could choose a tool like TypingMind, which allows you to connect AI assistants from many services, including OpenAI.

P.S.:

The article image at the top is not only generated by ChatGPT, but also based on its own idea. I had asked it how it want to be represented. It explains the visual as follows:

“When I imagine how I should be represented visually, I would not choose an ordinary robot, but a single conversational intelligence with many capabilities. The central, almost deity-like figure represents this one presence, while the many arms make visible the wide range of things I can do: write, research, analyze, code, generate images, work with data, or process language. I deliberately chose the strict symmetry and clear silhouette so that the image remains striking even at a small size, while revealing additional layers and details on a large screen.”