Models

Choose the model that fits your task — you can compare two models in the workspace (⚖ Compare two models).

Which model fits my task?

Briefly describe your question. Model Hub suggests suitable, currently available models.

AI-assisted · your task is sent only to GPT-5 nano for preselection

Your question or task

What matters most to you?

General information

Usage runs via the providers' official API interfaces — without training on your data — and is GDPR-compliant in combination with a European provider. 🇪🇺 Fully EU AI compliant: Mistral Large 3

Price classes (typical request: 1 page of question, 3 pages of answer): € = usually under 2 cents · €€ = approx. 2–15 cents · €€€ = 15 cents to over 1 euro. All cost figures are measured or estimated values in euros.

Conversations

Conversation: €0.00

⚖ Compare two models versus

≈ —

Projects

Flex account 1,000 Flex = €1.00 net · Flex is just the pricing unit — counting uses real provider tokens

—

Credit

—

Usage this month

—

Total top-ups

—

Receipt chain (tamper protection)

Counting rules — what costs Flex, and what doesn't?

Counting uses real provider tokens (usage fields of every response, taken over unchanged). Flex = tokens × published rate per model. Every receipt discloses the multiplication.
Thinking tokens ("reasoning work") are real provider output tokens and cost Flex — they are shown separately on the receipt, never hidden.
Cached content (cache) is billed at the cheaper cache rate — a visible benefit on the receipt.
Images cost a fixed price per image (per model/quality level); the automatic image captioning is billed as its own small line item.
Free for you: automatic subject assignment, project name suggestions, failed requests without a response, retries caused by KI-MARO S.A.
No rounding up per request — calculated exactly, displayed to one decimal; the monthly total is exact.
Cap guarantee: Before every request you see the range + cap. A running answer is never cut off — in borderline cases the balance may dip slightly below zero as a result (goodwill), and the next request then requires a top-up.
Rate changes only with advance notice and a change history; falling provider prices are passed on. Topped-up credit keeps its euro value (1,000 Flex = €1.00).

Price list in Flex

Flex per 1,000 model tokens. For comparison, the public provider list price and our markup factor — openly disclosed: the markup is our margin. Guide price = a typical request with 700 input and 2,100 output tokens. For providers with reported actual costs (research, Grok) the rule is: actual costs × factor. Source: public provider price lists, conversion 1 USD = €0.92.

Monthly overview

—

used (measured)

—

Monthly budget

—

Requests

—

Estimate quality (Ø actual/estimate)

All figures are our own measurements from the providers' response data × price list (as-of date below). Reconciliation with the providers' billing APIs is part of the full version — deviations due to exchange rates and daily cut-offs are possible.

Budget

Monthly budget € — notice at 70 %, per-request confirmation at 90 %, cheaper routes suggested from 100 %. It is never hard-blocked; every overrun is logged.

Memory is sent with every request as context — costs input tokens, visible in the estimate

Include memory

Operating mode the customer view calculates in Flex (customer prices); the operator view additionally shows purchase/list prices

Operator view (show purchase prices)

Purchase price list — operator view only

Euros per 1 million tokens (provider list prices = our purchase cost). Please check regularly against provider prices — the price list is deliberately editable and stored locally. Historical costs are not changed retroactively. The Flex customer rates are independent constants (a change = a new version with advance notice).

Sign in