Our price list, and how we set it
One price per model and per zone, 0% on top-ups. Here is where the numbers come from.
We are publishing our price list today. One price per model and per zone, per million tokens, excluding VAT, in euros.
The principle
For open-weight models, our price is the price of the European host that runs the model, plus 25%. For US models served by proxy (GPT, Claude, Gemini), it is the vendor's price plus 12%, invoiced in euros by a French company.
Top-ups are free of charge: €10 bought, €10 of credit. The price shown per model is the price paid.
What that gives
- gpt-oss-120b, EU zone: €0.10 input, €0.50 output, per million tokens. That is below Scaleway's public price for the same model.
- Mistral Large 3, EU zone: €0.56 / €1.67.
- Mistral Small 4, CH zone: €0.25 / €0.94.
The full list is on the pricing page.
Why a European price
Our prices are those of inference run in Europe. Open-weight models served in the United States cost two to ten times less: that is the gap between data you can locate and data whose destination you do not know.
Credits start at €10, the first key comes with €2 free, and credits stay valid for 12 months.