First the AI ​​wanted us to pay for its quality. Now we want us to pay for the right to use it without limits

Everything is invented, they say. The Freemium model It is a good proof of this: it is common for a long time that we can play video games or use free software applications. But of course, they are not the complete versions, but basic and capadas editions. So you can keep taking advantage of them if you don’t need anymore, but the opposite The problem beginsbecause it has to go through the box.

That is just what we are living With the world of AI. Almost all models allowed us to use them for free. It was what happened with Chatgpt, Gemini, Copilot or Claude.

Over time the companies responsible for developing these models were launching more powerful and higher quality versions, but if you wanted to use them, I had to pay. Chatgpt plus, Claude Pro either COPILOT PRO They marked the guideline with subscriptions of $ 20 per month that gave access to more intensive uses and also to those most advanced models and characteristics.

But that was not enough, and Openai announced the Chatgpt Pro launch In December 2024. It was an even more advanced and perfect alternative for intensive users, but its price also raised an important barrier, because not everyone is willing to pay $ a month for using AI models.

That plan may seem expensive if you do not take advantage of it, but what happens if you take out all the juice? It may then become an aid that allows you to do more in less time, and if so, those 200 dollars may seem to a bargain: it depends on each situation, and OpenAi knows it. So much so that he states that in the future there will be hyperspecialized chatbots that 20,000 euros will cost per month. It looks like an absurd figure, but what if you use them generate 40,000, 50,000 or 100,000 euros per month? There is the key: if you take advantage of those models intensively and if they really compensate.

And that’s where companies have realized that they can charge Not only for quality, but for the quantity.

This is what Anthropic also proposes, which these days announced his new Claude Max plans with two variants. The first, 90 euros per month, which allows limits of use five times higher than those of Claude Pro (the anthropic chatgpt plus, which costs 15 euros month). The second, 180 euros per month, which goes further and allows you to use this chatbot with more broad limits.

What does that mean? Basically, even being an intensive user it will be very difficult for the feared message to appear “you have exhausted the request of requests, wait a few hours to use it again.” It is not that the model is better or faster – at least, for now. You can simply use it much more.

And if you want more, nothing happens: there is the API

Anthropic’s strategy, such as Openai before, is logical and reasonable: if you want to “abuse” the advantages of AI and make intensive use of its capacity, you will have to pay an extra.

There is one more alternative in these areas, and that is that of Pay for the use of the API: Here we can consume as much as we want, and the payment for use is measured at cost per million input or output tokens. If the model is more modern and capable, you pay more. If it is older and modest, less.

In Anthropic, for example, Claude 3.7 Sonnet (its best model currently) costs three dollars per million entry tokens and $ 15 per million output tokens. The previous version, Claude 3.5 Haiku, costs $ 0.8 per million input tokens and $ 4 per million departure tokens.

With the payment model for use raised by the APIs, very directed to developers and companies – and especially useful for programmers – there are also great differences between some suppliers and others. The O1-PRO model of Chatgpt Pro is for example 140 times more expensive at its cost of API that the O3-mini that is available in Chatgpt.

What is clear is that companies have already profiled how we are going to pay to use artificial intelligence. We will not only pay for the one that is more advanced; If we want to use it a lot, it will touch more.

Here there is a danger that companies play with those limits and “Redoublacionen”but we will cross that bridge when we get to it. At the moment one thing is clear: using the AI ​​can leave for free, but if you want to use it a lot and well, go preparing the wallet.

Image | Solen Feyissa

In Xataka | The Ghibli paradox: the most viral success of AI is at the same time a symptom of its problems as a daily product