ChatGPT became available to the public on November 30, 2022

OpenAI released ChatGPT as a free preview on that date. The tool reached one million users in five days — faster than any consumer process at the time. This initial version was built on GPT-3.5, an earlier language model that OpenAI had developed.

The launch was not the start of the technology itself. OpenAI had been working on large language models for years before ChatGPT existed. What changed in November 2022 was that the company packaged one of these models into a chat interface and let anyone use it without special access or technical knowledge.

Before the public launch, ChatGPT had been tested by a smaller group of researchers and developers. OpenAI called this period the "research preview." The company used feedback from these early testers to fix problems and improve how the system worked before opening it to everyone.

Key Takeaways

  • ChatGPT launched publicly on November 30, 2022, as a free preview built on the GPT-3.5 model.
  • The technology behind ChatGPT existed before the launch, but the chat interface made it accessible to non-technical users.
  • OpenAI ran a research preview period with selected testers before the public release.
  • Major updates to ChatGPT have included the release of GPT-4 in March 2023 and the introduction of paid subscription tiers.
  • Version numbers and release dates matter because they affect which features you have and how the system behaves.

How ChatGPT's versions have changed since launch

The original November 2022 release used GPT-3.5, which was trained on text data up to early 2022. This version had clear limitations: it sometimes gave wrong answers with confidence, struggled with math, and could not access the internet or see images.

OpenAI released GPT-4 on March 14, 2023. This model was more capable across most tasks. It could understand images, handle longer text, and made fewer factual errors. However, GPT-4 was initially only available to paid subscribers. Free users continued to use GPT-3.5.

In September 2023, OpenAI made GPT-4 available to free users on a limited basis — the system would let free users access it a certain number of times per day before switching them back to GPT-3.5. This meant the version you saw depended on how many people were using the service at that moment.

Throughout 2024, OpenAI released smaller updates and refinements rather than major new versions. These included improvements to how the system followed instructions and handled specific types of tasks.

What the version numbers actually mean

When you see "GPT-3.5" or "GPT-4," the number refers to the underlying language model — the mathematical system that generates responses. A higher number generally means the model was trained on more data, trained for longer, or trained using improved methods.

The version you are using matters because different versions have different strengths. GPT-4 is better at reasoning through complex problems, understanding context in long documents, and avoiding false statements. GPT-3.5 is faster and uses less computing power, which is why OpenAI kept it available for free users even after GPT-4 existed.

The version also determines what the system knows. GPT-3.5 was trained on data through early 2022. GPT-4's training data extends further, though OpenAI has not published an exact cutoff date. Neither version can browse the internet in real time, so both will give outdated information about recent events.

Why OpenAI released ChatGPT as a preview, not a finished product

Calling the November 2022 release a "preview" was deliberate. OpenAI wanted to gather real-world feedback before committing to a final version. A research preview also gave the company legal and technical room to make changes without promising stability.

The preview label meant users should expect bugs, changes to how the system worked, and occasional outages. It also signaled that OpenAI was still learning what people actually wanted from the tool — information that would shape future versions.

Even after two years, OpenAI still describes ChatGPT as a research tool in some contexts. This framing protects the company from liability if the system makes mistakes, and it explains why features sometimes change or disappear without warning.

How updates and new versions are released

OpenAI does not announce every small change to ChatGPT. The company publishes major version updates (like the move from GPT-3.5 to GPT-4) and significant feature additions (like image understanding or file uploads). Smaller fixes and performance improvements often roll out without fanfare.

Updates are usually rolled out gradually rather than all at once. OpenAI might release a new feature to a small percentage of users first, watch how it performs, and then expand it to everyone. This approach reduces the risk that a broken update affects millions of people simultaneously.

You typically do not need to do anything to get an update. If you use ChatGPT through the web browser, you automatically get the latest version when you reload the page. If you use the mobile app, your phone's app store will offer an update, but you can usually delay installing it.

The difference between ChatGPT versions and OpenAI's other products

ChatGPT is the consumer-facing chat interface. Behind it sits the GPT model (GPT-3.5, GPT-4, and so on). OpenAI also sells access to these same models through an API — a technical interface for developers who want to build their own applications.

The API versions sometimes launch before or after the ChatGPT versions. For example, developers could use GPT-4 through the API weeks before it appeared in ChatGPT. This happens because the API serves different customers with different needs, and OpenAI manages their rollouts separately.

There is also GPT-4o (the "o" stands for "omni"), released in May 2024, which processes text, images, and audio. This version is available through both ChatGPT and the API, but again, the rollout timing and feature availability differ between the two.

Why the launch date matters for understanding how the tool works

Knowing that ChatGPT launched in November 2022 helps explain why it does not know about events after early 2022 (or later, depending on the version). It also explains why the system sometimes behaves inconsistently — it has been updated many times, and those updates sometimes change how it responds to the same question.

The launch date also matters for context. ChatGPT did not exist five years ago. It is a very new tool, which means there is still a lot that is unknown about how it will develop, what its limitations are, and how people will use it. When you read about ChatGPT's capabilities or problems, the date of the article matters — something written in 2023 may not reflect how the system works now.

Frequently Asked Questions

Is ChatGPT still in preview, or is it a finished product?

OpenAI still calls it a research preview in some official materials, but it functions as a finished product for most users. The preview label gives OpenAI flexibility to change features and pricing without the legal obligations that come with a "final" release. For practical purposes, treat it as a stable tool you can rely on, but understand that features and behavior can change.

Can I use an older version of ChatGPT if I prefer it?

Not directly. You cannot choose to use GPT-3.5 if you have access to GPT-4. However, free users automatically use GPT-3.5 (or limited GPT-4 access), while paid subscribers can select which model they want to use in the chat interface. If you prefer the older version's speed or behavior, a paid subscription gives you that choice.

What data was ChatGPT trained on?

GPT-3.5 was trained on text from the internet and other sources through early 2022. GPT-4 has a later training cutoff, though OpenAI has not published the exact date. Neither version can access the internet during conversations, so both will give outdated information about recent events. OpenAI has not released a detailed list of which websites or books were used.

Will there be a ChatGPT version 5?

OpenAI has not announced GPT-5 or a timeline for it. The company has released incremental improvements and specialized versions (like GPT-4o) rather than jumping to a new major version number. Future releases will depend on OpenAI's research progress and business decisions, which are not publicly predictable.

How often does ChatGPT get updated?

Major version releases (GPT-3.5 to GPT-4, for example) happen infrequently — months or years apart. Smaller updates, bug fixes, and feature additions happen much more often, sometimes weekly. You will not see most of these updates because they happen in the background. OpenAI publishes a changelog on its website if you want to track announced changes.