What are the new GPT-5 limitations compared to the GPT-4 family?

GPT-5 arrived a few weeks ago, though its rollout hasn’t been entirely smooth. While the model shows plenty of promise, its debut also meant the abrupt removal of every other GPT model from ChatGPT’s user-facing UI. Since then, some old models have returned, and there have been a few other changes to the way the system works. Furthermore, many of the initial GPT-5 usage limits have been temporarily enhanced since launch.

Let’s dive in and take a closer look at the GPT-5 limit structure and how it compares to what came before, as well as whether there’s any way to get back the older models.

New Limit Structure explained

When first launched, GPT-5 only had three tiers: GPT-5, GPT-5 Thinking, and GPT-5 Pro. This has since changed a bit.

Those with a free subscription to ChatGPT are limited to just GPT-5, which automatically switches to other GPT-5-based models, depending on the context. You are limited to just 10 messages every 5 hours, which makes it very easy to hit the limit now. Unlike Plus and Pro users, you also don’t have access to legacy models anymore.

Plus subscriptions get more options, with access to GPT-5 Fast, GPT-5 Thinking, and GPT-5 Thinking Mini. The exact limits vary as follows:

• GPT-5 (Auto): Virtually unlimited access, though it will often default to Fast for most basic inquiries. That said, if it automatically uses other models, these don’t count against your limits.
• GPT-5 Fast (Mini): Again, this model is virtually unlimited.
• GPT-5 Thinking: You can make 200 quests per day with GPT-5 Thinking, which is honestly much more generous than the old limits for the o3 Thinking model.
• GPT-5 Thinking Mini: You can make 2,800 requests a week with this model.

Only Pro users have access to the most powerful version of GPT-5, known as GPT-5 Pro. This tier also has virtually unlimited access to most tiers.

It’s worth mentioning that ChatGPT Plus actually launched with an 80 message per 3-hour limit, but OpenAI expanded the limit a few times since, though it is very likely some of these extensions are temporary. In other words, the limits are very much subject to change. Of course, that’s always kind of been the way with ChatGPT, as it’s older legacy models often saw limited revisions over time as well.

The good news is that even if you hit the wall, Plus users will still be able to use GPT-5; it will just automatically revert to GPT-5 Fast for most requests.

Here’s a quick refresher on Plus-tier GPT-4 limits:

• GPT-4o: 80 messages per 3 hours
• GPT-4: 40 messages per 3 hours
• GPT-4.5: 50 messages per week
• GPT-4.1: 80 messages per 3 hours
• GPT-4.1 mini: unlimited
• 03: 100 messages a week
• 04-mini-high: 100 messages per day
• 04-mini: 300 messages per day

Free users typically had 25–50 messages every 3 hours with GPT-4o. Pro subscribers enjoyed near-unlimited use across most models.

It was messy, but you could often work around limits by switching models. GPT-5’s unified approach is simpler, but it also removes that flexibility.

What is each GPT-5 sub-model best used for?

While ChatGPT officially breaks down its limits and capabilities, the truth is that official differences are vaguely stated at best. That said, I’ve been using GPT-5 quite a bit since launch, so here’s a general breakdown based on my experience:

• GPT-5: This is the best for low-stakes answers, as it will often relegate itself to the Fast/Mini model for many basic questions. The answers are often decent, but this tier needs the most independent confirmation and research before assuming its right.
• GPT-5 Thinking: This is best for deeper data analysis, scientific questions, philosophy, hypotheticals, and other scenarios where a deep chain of thoughts is needed to get a solid answer. Even then, it’s still important to double-check the work, but it hallucinates a lot less than ever before.
• GPT-5 Thinking Mini: Like Thinking, this is a good model for research and searching in general, where you want thought put in but you don’t need to go beyond the surface. For example, maybe you want a list of research papers on a specific topic. You’ll still need to verify your work, but this is a fast way to find good sources for essays, articles, and other creative or educational projects.
• GPT-5 Pro: This model is for heavier research, analysis, and other professional projects and requires a Pro-tier subscription.

If you’re on ChatGPT Pro, yes — you still have the GPT-4 family available to you. Free users do not. Plus subscribers officially see only GPT-5 and GPT-5 Thinking, but there’s a partial workaround. Head to Settings > General and enable Show Legacy Models. This restores GPT-4o at least, which could be handy if GPT-5’s limits or style aren’t working for you.

For reference, GPT-4o still offers around 80 messages every 3 hours for Plus users. This is handy if you’ve not been overly impressed with the rollout of GPT-5.