what each model costs

AI models side by side, with the source included.

Every price below comes word for word from the provider’s own page, with the date on which we checked it. We convert nothing and estimate nothing: what cannot be checked at the source does not go in, and then the reason stands where an amount would be.

As of 29 August 2026. Prices and models change quickly; check the provider’s official page before you pay.

What the providers cost per month

Four providers, three questions: can you try it for free, what does the cheapest plan cost, and what does the next step up cost. Every price is shown in the currency and the form in which the provider itself shows it to a visitor from the Netherlands. We convert nothing and add nothing up, because a converted or recalculated amount is our number and not the provider’s. Watch the tax: Anthropic states that its amounts exclude tax; the other pages say nothing about it.

Prices per provider, with source and check date per cell
ProviderFree to tryCan you work with it without paying?Entry planThe cheapest paid plan per month.Next step upWhat it costs if you need more usage.Checked
ClaudeAnthropicFree to tryYes, €0 per monthsourceEntry plan€18 per month€15 per month with annual billing, €180 up front; the page states that the amounts exclude taxsourceNext step upfrom €90 per monthMax; within it you choose 5x or 20x more usage than Prosource
GeminiGoogleFree to tryYes, €0 per monthsourceEntry plan€4.99 per monthGoogle AI PlussourceNext step up€21.99 per monthGoogle AI Pro; above that, Ultra from €99.99source
ChatGPTOpenAIFree to tryYes, €0 per monthsourceEntry plan€8 per monthGo; the page states that this plan may contain adssourceNext step up€23 per monthPlus; above that, Pro from €103source
LovableLovableFree to tryYes, $0 per monthsourceEntry plan$25 per monthPro, with 100 credits per monthsourceNext step up$50 per monthBusiness, with 100 credits per monthsource

What each provider is meant for

Price is rarely the reason someone switches. Where it gets stuck is that the model is good at something other than what you use it for. Per provider, one line on what it aims at, without judgement.

ClaudeAnthropic

Writing, reasoning and code. Also the engine behind some of the build tools.

As of 29 August 2026. Prices and models change quickly; check the provider’s official page before you pay.

GeminiGoogle

Writing and search, linked to the rest of Google.

As of 29 August 2026. Prices and models change quickly; check the provider’s official page before you pay.

ChatGPTOpenAI

The most widely known model, often the first one people try.

As of 29 August 2026. Prices and models change quickly; check the provider’s official page before you pay.

LovableLovable

A build tool: you describe what you want and a working app comes out.

As of 29 August 2026. Prices and models change quickly; check the provider’s official page before you pay.

How to choose when you are just starting

Start with the free tier of two providers and use them for a week for the work you actually do, not for a test question. The difference you notice then is bigger and more honest than any comparison overview, this one included.

Only pay after that, and pay monthly until you are sure you will stay. With annual billing the monthly price is lower, and that is exactly enough to lock you into a choice you could not yet make after three weeks.

With a build tool, look at something other than with a chat model: not the price per month but whether you can get your project out. A tool you can export from costs you an afternoon when you switch; a tool where that is not possible costs you the project.

Building something yourself with an API?

Then you pay per token and not per month, and this is the wrong table. Fifteen models with their price per million tokens and their score on an independent index are on a separate page, with the same source and date per cell as here.

What this table does not say

A price table is easy to read as a ranking, and it is not one. Three things are deliberately left out, because they cannot be backed by a source.

  • No judgement of qualityWhich model answers better depends on what you use it for. That does not fit in a cell, and attaching a number to it would be an opinion that looks like a fact.
  • No converted amountsThree providers show euros and Lovable shows dollars. We do not convert, because then today’s exchange rate would sit in a cell that is no longer right tomorrow and that the provider never published in that form.
  • No comparison of what you get per planOne provider counts messages, another credits, a third talks about more usage without a number. Those units cannot be converted into each other, so only the price is shown and not what you get for it.

What is shown can be checked per cell: click the source and you land on the page the amount comes from, with the date on which we last looked.

Further reading

Once you have chosen your tool, the real work only begins. These three are about what comes after.

From first prompt to tax return.

basestep is not in this table and is not compared against anything here.