Intelligence: AI chat, pay as you go
Chat with current AI models such as Gemma, Mistral, GPT-OSS or Kimi in the customer portal. You only pay for what you use and decide yourself how much it may cost per month at most.
Then activate it in the customer portal under “Intelligence”.
What you get
AI models in the customer portal, with no base fee and full cost control.
-
Only pay for what you use
You pay per token, that is per processed piece of text. One invoice per month, no base fee.
-
Full cost control
You set your own monthly limit and see your usage live in the portal. At 80% we let you know.
-
Eight models, one chat
From fast and affordable to thorough: you choose the model for each message.
-
Your inputs belong to you
According to Cloudflare, your inputs are not used for training. You can delete your history at any time; otherwise it is deleted automatically (default: after 90 days without activity).
How it works
Four steps to your first chat.
- Create a customer accountSign up in the customer portal and confirm your email address. Already a customer? Simply log in.
- Add a payment methodCard (credit or debit card), SEPA Direct Debit or Link by Stripe, plus your billing address. Both are stored securely with Stripe.
- Activate IntelligenceUnder “Intelligence” in the customer portal, you choose your monthly limit and confirm the terms.
- Get startedStart chatting right away. We bill monthly in arrears.
Models and prices
Prices in euros including statutory VAT per 1 million tokens, as of 1 October 2026.
| Model | Input | Output | Cached input |
|---|---|---|---|
| Gemma 4 26B DefaultGoogleFast and affordable: everyday questions, writing, summaries. | €0.16 | €0.48 | no separate price, the input price applies |
| GLM 4.7 FlashZ.aiVery affordable, many languages. | €0.10 | €0.63 | no separate price, the input price applies |
| Mistral Small 3.1 24BMistral AIModel from a European maker, balanced for writing and questions. | €0.56 | €0.88 | no separate price, the input price applies |
| GPT-OSS 120BOpenAIStrong open model for more demanding tasks. | €0.56 | €1.18 | no separate price, the input price applies |
| Nemotron 3 120BNVIDIACareful reasoning for complex questions. | €0.79 | €2.36 | no separate price, the input price applies |
| DeepSeek V4 Flash, may be temporarily busy when demand is highDeepSeekLogic, maths and coding. | €0.70 | €2.08 | €0.03 |
| GLM 5.3 Flash, may be temporarily busy when demand is highZ.aiCoding and analysis at a good price. | €0.24 | €0.79 | €0.05 |
| Kimi K2.6, may be temporarily busy when demand is highMoonshot AIPremium all-rounder for demanding tasks. | €1.50 | €6.29 | €0.26 |
The prices shown in the customer portal apply. ¹ May be temporarily busy when demand is high. – No separate price for cached input; the input price applies.
Licence and model information: Gemma 4 26B, GLM 4.7 Flash, Mistral Small 3.1 24B, GPT-OSS 120B, Nemotron 3 120B, DeepSeek V4 Flash, GLM 5.3 Flash, Kimi K2.6.
What is a token?
A token is a piece of a word, roughly three to four characters of text.
A typical question with an answer (around 2,000 input tokens and 500 output tokens) costs about 0.06 cents with Gemma 4 26B and about 0.61 cents with Kimi K2.6.
With every question, the conversation so far is sent along as input, so long conversations cost more. Some models “think” before answering; these thinking tokens count as output.
How we bill
You know in advance what you pay per token and decide yourself how much it may be per month.
- Monthly in arrearsWe bill once a month via Stripe and collect the amount from the payment method you have stored.
- Small amounts carried forwardAmounts below €5 are carried forward to the following month, for at most 12 months.
- Monthly limitDefault €10, from €1, initially up to €25, up to €100 after your first paid invoice, more on request. Once it is reached, Intelligence accepts no further requests until the end of the month.
- No base fee, no minimum termYou can cancel at any time in the customer portal.
All prices in euros, including statutory VAT. Intelligence is billed by usage at the token prices above.
Payment method required
Because we bill by usage, you need a stored payment method (card (credit or debit card), SEPA Direct Debit or Link by Stripe) and a billing address before you start. You store both securely with our payment service provider Stripe.
Data protection and AI notice
The models run on Cloudflare (Workers AI). The place of processing is not limited to the EU and may also be in the USA (EU-US Data Privacy Framework, Standard Contractual Clauses).
According to Cloudflare, your inputs are not used to train models; the model makers receive no data.
We store your conversations in the customer portal until you delete them or they are deleted automatically (default: 90 days without activity, and 14 days after the end of the contract). We store at most 1,000 conversations per customer; beyond that, we automatically delete the oldest ones. For billing, we only store the time, model, tokens and amount, no content.
Do not enter particularly sensitive data, passwords or payment details.
Details are in our privacy policy and in § 21 AGB (German).
Frequently asked questions
Which models are available?
Currently 8: Gemma 4 26B (default), GLM 4.7 Flash, Mistral Small 3.1 24B, GPT-OSS 120B, Nemotron 3 120B, DeepSeek V4 Flash, GLM 5.3 Flash and Kimi K2.6. You choose the model for each message. The customer portal shows which models are available; we may add or replace models.
What does a message cost?
That depends on the model and the length. A typical question with an answer (around 2,000 input tokens and 500 output tokens) costs about 0.06 cents with Gemma 4 26B. Long conversations cost more, because the conversation so far is sent along as input. After each answer, the chat shows what it cost.
What happens when my limit is reached?
Intelligence then accepts no further requests until the end of the month; we never charge more than the limit that applied at the time of each request. You get an email at 80% and at 100%. You can change the limit in the customer portal at any time, initially up to €25 and up to €100 after your first paid invoice.
Are my inputs used for training?
No. According to Cloudflare, inputs and outputs are not used to train models and are not stored by Workers AI. The model makers receive no data. We do not look at your conversations unless you ask us to, or it is necessary to investigate a specific suspicion of misuse or because of a legal obligation.
Where is my data processed?
The models run on Cloudflare (Workers AI); the place of processing is not limited to the EU and may also be in the USA. The transfer is based on the EU-US Data Privacy Framework and the EU Standard Contractual Clauses. The customer portal stores your conversations in its database at Cloudflare (primary location Western Europe). Details are in our privacy policy.
Do I need a hosting plan?
No, just a customer account. You can use Intelligence independently of our other products.
Is there an API?
Planned, but not yet available. For now, you use Intelligence via the chat in the customer portal.