You Probably Don't Need to Pay for ChatGPT Anymore
The big tech story last week was GPT-4o, the newest model from OpenAI. GPT-4o is more than a simple ChatGPT upgrade, however. The headlining feature is the new version of Voice Mode powered by GPT-4o, which essentially turns ChatGPT into Samantha, the smart assistant from 2013's Her (even if "she" no longer sounds like ScarJo). While the demos showing off Voice Mode understandably went viral, one other piece of GPT-4o news was at least as significant: The newest model, including the futuristic and dystopian Voice Mode, will be offered to all users for free.
This is a first for OpenAI. While you've always been able to access some ChatGPT services for free, paying for ChatGPT Plus upgraded the experience considerably. In the past, the subscription came with access to the latest and greatest models that free users couldn't access, as well as other perks, such as ChatGPT Plugins and their successor, GPTs.
Now, however, OpenAI seems to be leveling the playing field for everyone, offering many of Plus' formerly exclusive features to free users.It feels like there must be a catch—and there kind of is—but even still, free users have never been able to access so much ChatGPT before, to the point that it's probably not worth paying $20 for ChatGPT Plus anymore.
What a free ChatGPT account gets you
Whether you make a free ChatGPT account, or log in with the same account you've had since 2022, you now have (or will soon have) access to the GPT-4o model. It may be enabled by default, or you may need to switch to it manually. (On mobile or desktop, select the model name at the top of the screen (i.e. ChatGPT 3.5) and make sure GPT-4o is selected.)
Once selected, you have access to the same GPT-4o model that paid users do. OpenAI says it's better than previous models at understanding the images you upload, and will have improved responses and insights in turn. In addition, while you should see about the same performance when dealing with English text with GPT-4o versus previous models, OpenAI says it's much better at processing non-English texts. (Check out some anecdotal evidence to see how much faster GPT-4o can be in comparison to GPT-4.)
More features and faster performance
This is a big deal, since the previous models, GPT-4 and GPT-4 Turbo, were not available in the free tier. (That said, you could previously access GPT-4 Turbo using Copilot.) But it'll be an ever bigger deal down the line, once OpenAI does roll out the new GPT-4o Voice Mode—while Plus users will get Voice Mode first, free users will also receive it in the near future.
But even before Voice Mode gets here, free users have a slew of new features to try. With the new model, you now have access to data analysis, file uploads, vision, and web browsing, features tied to GPT-4o. GPT-3.5 can't read uploads, view photos and videos, or search the web, so this is the first time these options are available to users who don't pay for OpenAI's services.
In addition, the free plan now comes with access to GPTs and the GPT Store. GPTs are custom-built versions of ChatGPT that specialize in a particular skill or service. This feature was previously a Plus-exclusive, but now free users can browse the GPT store for bots that specialize in everything from translation to meal-prepping. Plus, they can make their own GPTs if they don't find what they're looking for. Memory is another paid exclusive turned free—now, ChatGPT will remember past conversations for free users too, so a new chat will no longer be the total blank slate it used to be.
Why pay for ChatGPT at all?
If you're really serious about GPT-4o, the answer is simple: message limits. OpenAI says that paid users will have five times the message limits for GPT-4o than free users, but isn't specific about what the message limits actually are for either plan. If Reddit is to be believed, however, free users may only have anywhere from 10 to 16 messages every three hours using GPT-4o. That's not much, especially if you frequently chat with OpenAI's chatbot. Once you hit that limit, you'll need to drop back down to the GPT-3.5 model, which isn't bad, but will likely feel much more limited compared to the latest model.
But it's not just better performance you lose out on during these periods of downtime: GPT-4o is what gives free users access to data analysis, file uploads, vision, and web browsing. When you run out of messages with GPT-4o, you also run out of chances to use these features, until your message limit resets in three hours.
That's not to say that paid users have unlimited access to these features: As of May 13th of this year, Plus can send 80 messages every three hours on GPT-4o, and 40 messages every three hours on GPT-4. (Why the discrepancy? GPT-4o is actually 50% cheaper than GPT-4.) While it's not like paying $20 per month will give you uninterrupted access to GPT-4o and GPT-4, you certainly will have more time with these models and their features than free users.
That will likely apply to the new Voice Mode when it drops, as well. You may find yourself running out of messages quickly while chatting with the model on the free tier, while paid users can extend their conversations by perhaps as much as 800%.
Most users will b fine with the free tier
In a way, this new approach from open AI adds a sort of "ChatGPT Plus demo" to the free tier: At any time, you can switch over to the best that OpenAI has to offer, and see what life is like on the other side. If you only use ChatGPT for short bursts, this may work for you. You can ask a question that requires a web search, upload a photo and ask ChatGPT something about it, and have the model run an analysis on a data set for you, then close out of the app without having paid a dime. But if you need to do these things frequently enough that you hit those message limits, you may be tempted to reach for your credit card.
In my view, though, most of us will be able to get by without having to pay for ChatGPT anymore: For most simple tasks, GPT-3.5 is still useful enough, while GPT-4o can help with more complicated or multimodal requests. Plus, if you really need a faster GPT model in a pinch, there's always Copilot.