On this page
Most people need one subscription, not three. That is the honest headline, and it saves more money than any comparison below. If you genuinely do need several models, there are three routes, and which one wins is arithmetic rather than opinion.
Prices below were checked on 23 September 2026. They change; check before you buy.
Do you actually need more than one?
Start here, because it is where the money is.
Two general chat subscriptions costs roughly €40 a month. Almost nobody acts on the second opinion often enough to justify it — you ask the same question twice, notice the answers differ in tone rather than substance, and go with the first one anyway.
The test that settles it: cancel one for a month. Keep a note each time you genuinely wanted the other. If that note has fewer than five entries, you have your answer, and you are €240 a year better off.
Where a second model earns its place is when the two are doing different jobs — one for writing and reasoning, one for code, or one with access to your files. That is a real reason. "In case" is not.
What are the three routes?
| Route | What you pay | Suits |
|---|---|---|
| One direct subscription | ~€20/mo | almost everyone |
| An aggregator | one fee, many models | people replacing two or more subscriptions |
| Pay per use (API) | per token | occasional heavy bursts |
Direct
Claude is about €20 a month, and the equivalents from other vendors sit in the same range. You get the newest model the day it arrives, the official app, and no middle layer between you and the vendor.
That last point matters more than people weigh it. With an aggregator, whether you get the newest model — and how quickly — is the aggregator's decision, not yours.
Aggregators
One subscription, many models behind it. GlobalGPT and Lorka both work this way.
GlobalGPT's published tiers, billed annually, run at $5.80, $10.80 and $25.00 per month at the time of writing, with the top tier advertising access to the full set of models. Lorka sits at roughly $30 a month and bundles writing, images, video and voice rather than chat models alone.
Against two direct subscriptions at €40, either can win. Against one at €20, neither does unless you actually use the breadth.
Pay per use
Z.ai (GLM) and similar services charge for what you consume instead of monthly. This is the right shape for bursts — a week of heavy work, then nothing — and the wrong shape for daily chat, where a subscription is effectively unlimited and per-token spending is hard to predict until you have a month of it behind you.
I have not published per-token figures here because the ones I could verify are not on a page I can point you at with confidence. Price the specific model you would use, at your expected volume, before switching anything.
The whole list
68 tools, what each costs and when you actually need it. Updated twice a year.
Why the savings on their pricing pages are nonsense
Worth knowing, because it is the number that sells these products.
GlobalGPT's page advertises saving $1,233 a month against $1,282 of separate subscriptions. That figure is arrived at by adding up the list price of every service in the bundle — chat models, image tools, video tools, the lot.
Nobody buys all of that. The saving is measured against a person who does not exist.
This is not an argument against the product. It is an argument against the comparison. Your saving is: what you pay today for the things this would actually replace, minus the subscription. For most people that sum is either "€40 becomes $25, so yes" or "€20 becomes $25, so no" — a long way from $1,233 either way.
Which calculation actually decides it?
Write down what you pay now, then strike out anything the aggregator does not genuinely replace:
If the answer is negative, you are buying breadth, not savings. That can be a fine reason — just be clear which one you are doing, because the pricing page is designed to blur it.
The failure mode is stacking. You subscribe to the aggregator, keep your existing subscription because you are used to it, and now pay €45 a month for what used to cost €20.
What to check before you buy an aggregator
Four things, and the first two are where people get caught:
- Which models are included at the tier you would actually buy — not the tier in the headline. Cheaper tiers frequently exclude the newest models, which are usually the reason you wanted access.
- How quickly new models arrive. Direct means launch day. Through a middle layer it can be weeks, and sometimes never for the top tier of a model.
- Whether it replaces at least two things. One replacement almost never justifies the switch.
- What happens to your history if you leave. Conversations inside an aggregator are usually not portable.
What I use, and what it costs
One subscription and one coding agent: Claude at around €20 and Claude Code at around €17–20. About €40 a month, and that is the whole model budget.
I have tried running two general subscriptions twice. Both times I cancelled the second within two months, and both times the reason was the same: I was not comparing answers, I was just using whichever tab was already open.
The full breakdown of everything else I pay for is in what AI tools actually cost per month, and the side-by-side of the access routes is in how to pay for AI models.
The Short Version
Every conclusion from all eight playbooks — 97 chapters, condensed. Free.
Tools in this piece
| Tool | What it does | Price |
|---|---|---|
| Claude | The assistant itself. Strong on long text, code, and following instructions exactly. | ~€20/mo |
| GlobalGPT | Chat, image, video, audio, search and agents from several leading models under one subscription. | From ~$5.80/mo (yearly) |
| Lorka | One subscription covering writing, images, video and voice instead of five. | ~$30/mo |
| Z.ai (GLM) | Access to the GLM model family with credits for developers and builders. | Credits, pay as you go |
| Pi | AI presentation generator — paste the content, get a designed deck back. | Free tier, then paid |
| Claude Code | An AI agent in your terminal that reads your files, edits them, runs commands and deploys. | ~€17–20/mo |
Every tool has its own page with the price, who should skip it and what to check before paying. Some links are affiliate links.
Questions people ask
What is the cheapest way to access multiple AI models?
For most people, one direct subscription at around €20 a month, because the second model gets used far less than expected. If you genuinely need several, an aggregator subscription costs less than two direct ones — but only when it replaces them rather than being added alongside them.
Are AI model aggregators worth it?
They are worth it when they replace at least two subscriptions you already pay for, and when the models you actually use are included at the tier you would buy. They are not worth it as a third subscription on top of the two you keep renewing out of habit.
Why do aggregators claim savings of hundreds per month?
Because they add up the list price of every service they bundle, including ones nobody subscribes to. One aggregator advertises saving $1,233 a month against $1,282 of separate subscriptions. That comparison is only meaningful for someone who would otherwise buy all of it, and nobody does.
Is the API cheaper than a subscription?
For occasional heavy use, often yes, because you pay per token instead of per month. For daily chat use it is usually worse, because a subscription is effectively unlimited and per-token costs are hard to predict. The switch point is volume, not preference.
Do you need both ChatGPT and Claude?
Rarely. Two general subscriptions is the most common overlap in this category, about €40 a month for a second opinion most people do not act on. Try one month with only one and see whether you actually reach for the other.