ChatGPT Plus costs €21.02 per month, Claude Pro costs €21.02 per month and Google AI Pro costs €21.99. Less than a euro apart, and yet you're buying three fundamentally different things. With OpenAI you buy the broadest set of features, with Anthropic the best language model with a programming environment around it, and with Google really a storage subscription that happens to include a strong model.
So the price doesn't help you choose. What does help: looking at which models you get, how much you're allowed to do with them, what else is included and how each provider handles your data. This article puts those four points side by side and ends with a verdict for each type of user.
What exactly do they cost, per month and per year?
| Subscription | List price | Per month incl. VAT | Per year |
|---|---|---|---|
| ChatGPT Plus | $20 | €21.02 | €252.24 |
| Claude Pro (monthly) | $20 | €21.02 | €252.24 |
| Claude Pro (annual billing) | $200 per year | €17.52 | €210.23 |
| Google AI Pro | €21.99 | €21.99 | €263.88 |
One difference is administrative, but it counts. OpenAI and Anthropic bill in dollars excluding VAT, so your monthly cost moves with the dollar exchange rate and 21% VAT is added on top of the $20. Google prices its Dutch consumer subscriptions directly in euros including VAT, so €21.99 stays €21.99. Of these three, only Claude offers a real annual discount: $200 instead of 12 × $20 saves €42 per year. All amounts and the conversion are worked out in what does an AI subscription cost.
| Provider | Plan | List price | Per month in € | Per year |
|---|---|---|---|---|
| GoedkopeAIAll models below | Plus | €12.95 incl. VAT | €12.95 | €155 |
| ChatGPTOpenAI | Plus | $20 excl. VAT | €21.02 | €252 |
| ClaudeAnthropic | Pro | $20 excl. VAT | €21.02 | €252 |
| GeminiGoogle | Google AI Pro | €21.99 incl. VAT | €21.99 | €264 |
Which models do you get for your money?
| Subscription | Top model | Released | Context window |
|---|---|---|---|
| ChatGPT Plus | GPT-5.6 Sol | 9 July 2026 | 1.05 million tokens |
| Claude Pro | Claude Opus 5 | June 2026 | 1 million tokens |
| Google AI Pro | Gemini 3.1 Pro | preview | 1 million tokens |
On paper they're close: all three around a million tokens of context, all three a 2026 generation. In practice you mainly notice the differences in character. On 6 August 2026 GPT-5.6 Sol got an update with better factual accuracy and an effort slider that lets you choose how much thinking the model puts in. Claude Opus 5 is built for heavy reasoning, and Claude Sonnet 5 does the everyday work. Gemini 3.1 Pro is the heavy model in both Google AI Plus and AI Pro, with Gemini 3.5 Flash as the stable workhorse below it.
One thing is important to know before you take Claude Pro: Anthropic's very best model, Claude Fable 5, is not included in Pro. In the apps you pay for Fable 5 with prepaid credits. Anyone who thinks €21.02 gets them access to Anthropic's top model will be disappointed.
| Model | From | Released | Context | Dutch | Best for |
|---|---|---|---|---|---|
| GPT-5.6 Sol | OpenAI | 9 July 2026 | 1,05M tokens | Good | Heavy analysis, agents and professional work |
| Claude Opus 5 | Anthropic | June 2026 | 1M tokens | Excellent | Complex reasoning tasks and long documents |
| Gemini 3.1 Pro | preview | 1M tokens | Good | Research and document analysis |
How much are you allowed to use?
Unlimited doesn't exist with any of the three, and the way they limit usage differs.
ChatGPT is the most transparent: Plus gives you 160 messages per three hours on the top model. That's a limit you rarely hit in normal work, but you will in a long session of document analysis.
Claude works with usage windows instead of a message counter. That's harder to estimate, and with long documents or large codebases you run into it sooner than it feels like you should. On the other hand, Claude Sonnet 5 is also in the free tier, so you can test fairly well how hard you'll hit limits before you pay.
Google AI Pro doesn't sell a number of messages but capacity: the context window of one million tokens is the key measure. For anyone who wants a stack of PDFs read at once, that's the more relevant limit.
What else is included in the subscription?
This is where the three diverge the most, and this is usually what decides the choice.
| Subscription | What else is included |
|---|---|
| ChatGPT Plus | Sora 2 for video, GPT Image 2, Codex, Agent Mode, Deep Research |
| Claude Pro | Claude Code, Claude Cowork |
| Google AI Pro | NotebookLM, Deep Research, 2 TB storage, integration with Gmail and Drive |
ChatGPT Plus is by far the broadest package. Sora 2 and GPT Image 2 aren't in any subscription in this price range from another provider, and Agent Mode plus Codex make it more than a chat window.
Claude Pro has the narrowest package but the most significant component if you program: Claude Code is a full programming environment and is simply included in the €21.02. For a developer, that actually makes Claude Pro the cheapest of the three. For the comparison with GitHub Copilot Pro ($10) and Cursor ($20), see coding with AI.
Google AI Pro is the least an AI subscription and the most a package. You get 2 TB of storage, which is worth money in itself if you need that storage anyway, plus NotebookLM for working with your own sources. If you're not in the Google ecosystem, you're paying for something you leave untouched. For comparison: the cheaper Google AI Plus costs €7.99 and gives you 200 GB of storage.
What about your data?
Of the three, Claude Pro is the clearest: on paid consumer subscriptions, training on your conversations is off by default. If you turn training on yourself, the retention period actually becomes longer — a trade-off Anthropic makes explicit instead of hiding it.
With ChatGPT Plus and Google AI Pro you have to check your settings yourself; for these two we don't have a default setting we can firmly confirm, and then it's better to claim nothing than to suggest something.
What is certain: all three are American providers. If you want processing within the EU, none of these three is the answer. Then you end up with Mistral's Le Chat Pro (€15.76), which processes in France and Sweden and has a no-telemetry mode in the Pro subscription. For business users there's also Microsoft's EU Data Boundary. The full trade-off is in AI and privacy in the Netherlands.
Which one writes the best Dutch?
This is the criterion missing from English-language comparisons, and for Dutch users it's often decisive. From our own tests, on a scale of 1 to 5:
| Provider | Score | Verdict |
|---|---|---|
| Claude | 5 | The most natural Dutch: correct idiom, no anglicisms, tone you can steer |
| ChatGPT | 4 | Grammatically fine, but American in tone with the occasional literal translation |
| Gemini | 4 | Correct and good at translating, but flat and businesslike |
The difference between a 4 and a 5 sounds small and in practice isn't: with a 4 you're still polishing sentences that are structured just a little too much like English after every text. If you write Dutch professionally, Claude Pro pays for itself in editing time. The extensive test is in which AI speaks the best Dutch.
Which one should you choose?
- You write a lot of Dutch texts → Claude Pro. Language quality is the only point on which one of the three really pulls ahead of the rest.
- You program → Claude Pro, for Claude Code. If you'd rather stay in your editor, GitHub Copilot Pro at €10.51 is cheaper than all three.
- You work with documents, reports and research → Google AI Pro, for the context window, NotebookLM and Deep Research. Especially if you already needed that 2 TB of storage.
- You want images, video or voice → ChatGPT Plus. No competitor at this price level has Sora 2 and GPT Image 2.
- You don't yet know what you need → start free. Claude gives you Sonnet 5 without paying, and since 6 August 2026 ChatGPT gives you unlimited text chat on GPT-5.6 Luna. After a month you'll know where you run into limits.
- You need all three → then €64.03 per month, or €768.36 per year, is the price of that honesty. Whether that's sensible, we calculate in why multiple AI subscriptions are a waste of money, and the full field is in the comparison of all AI subscriptions.
Conclusion
Frequently asked questions
Which is better: ChatGPT Plus or Claude Pro?
For images, video and voice, ChatGPT Plus is clearly better: Sora 2 and GPT Image 2 aren't in either of the other two. For writing in Dutch and for programming, Claude Pro wins, because Claude is the only one to score a 5 in our language test and Claude Code is included. Both cost €21.02 per month.
Is Google AI Pro worth €21.99?
It is if you work with lots of documents and already use Google Drive, Gmail and Photos. You get Gemini 3.1 Pro with a context window of one million tokens, Deep Research, NotebookLM and 2 TB of storage. If you don't use that storage, you're paying for part of the package you leave untouched.
Which of the three writes the best Dutch?
Claude. In our own test Claude scores a 5 on Dutch, with correct idiom and a tone you can steer. ChatGPT and Gemini both score a 4: grammatically fine, but ChatGPT sounds American and Gemini flat and businesslike.
Which of the three is best for privacy?
Claude Pro has the clearest default setting: training is off by default on paid consumer subscriptions, and if you turn training on, the retention period actually becomes longer. If you want European processing, none of these three is the answer and you end up with Mistral's Le Chat Pro, which processes in France and Sweden.
Can you combine the three for less money?
Together, ChatGPT Plus, Claude Pro and Google AI Pro cost €64.03 per month, or €768.36 per year. A bundle subscription that brings together the models of several providers is cheaper, but lacks the provider-specific extras such as Sora 2, NotebookLM and 2 TB of storage.