Billing

Per token

Each successful call is paid from your prepaid credit, to the token: prompt_tokens at the model's input price, completion_tokens — reasoning included — at its output price. Prices are per million tokens, VAT included, so the credit you add is what you can spend (prices per model).

Price per million tokens, VAT included

Model	ID	Input	Output	Runs in
Mistral Large 4	mistral-large-4	€2.5133	€7.7246	EU region of Mistral AI
DeepSeek V4.1 Flash	deepseek-v4.1-flash	€1.008	€3.024	Germany (AKI.IO)
GLM-5.3	glm-5.3	€3.024	€9.24	Germany (AKI.IO)
Qwen3.8 27B	qwen3.8-27b	€1.008	€5.544	Paris, France
DeepSeek V4 Flash	deepseek-v4-flash	€0.672	€1.344	Paris, France
Apertus v1.5 70B	apertus-v1.5-70b	€1.2936	€4.62	Switzerland (Infomaniak)
GLM-5.2	glm-5.2	€3.024	€9.24	Paris, France
Qwen3.6 35B	qwen3.6-35b	€0.42	€2.52	Paris, France
Gemma 4 26B	gemma-4-26b	€0.168	€0.84	Germany (AKI.IO)
Mistral Small 4	mistral-small-4-119b	€0.2772	€1.1088	EU region of Mistral AI
Qwen3.5 397B	qwen3.5-397b	€1.008	€6.048	Paris, France
GPT-OSS 120B	gpt-oss-120b	€0.252	€1.008	Paris, France
Qwen3 Coder 30B	qwen3-coder-30b	€0.336	€1.344	Paris, France
Qwen3 235B	qwen3-235b	€1.26	€3.78	Paris, France
Mistral Small 3.2	mistral-small-3.2	€0.252	€0.588	Paris, France
Llama 3.3 70B	llama-3.3-70b	€1.512	€1.512	Paris, France
Pixtral 12B	pixtral-12b	€0.336	€0.336	Paris, France

When the operator serves part of the prompt from its cache and reports it in usage.prompt_tokens_details.cached_tokens, those tokens cost less on the following models (price per million cached tokens): Mistral Large 4 (€0.2587 instead of €2.5133), DeepSeek V4.1 Flash (€0.2016 instead of €1.008), GLM-5.3 (€0.42 instead of €3.024), Qwen3.8 27B (€0.2016 instead of €1.008) and DeepSeek V4 Flash (€0.1344 instead of €0.672). On every other model, the whole prompt is paid at the input price.

Embeddings

Embeddings use input tokens only, billed at the model's input price — there is no output price (Embeddings).

Images, per second

Images are the exception: they are paid per second of generation, from the same credit, at €0.0092 a second, VAT included. Evanesca measures the call to Infomaniak and rounds it up to the next second, one second at least; usage.generation_seconds is exactly what is billed. As a guide, about €0.02 for a FLUX.1 [schnell] image and about €0.05 for a PhotoMaker V2 portrait (Images).

Web search, per call

Over MCP, web_search and web_fetch are billed per call, VAT included: €0.0076 a search, €0.0015 a page — about €7.56 for 1,000 searches. That is all you pay for them: your client's own model reads the results (Web search).

Credit and invoices

There is no subscription: you add credit from €2 and it never expires (pricing). An invoice is sent automatically with every credit purchase: a PDF by e-mail, with VAT shown.

When credit runs out

When your credit reaches zero, new calls are refused with 402. Calls already running at that moment are charged in full once the operator reports their usage, so the balance can go slightly below zero; the difference comes off your next credit purchase.

Your account's record

Your account lists every call with its tokens and its price. A streamed call is charged from its final usage chunk — including when your client hangs up early, since the operator finishes the answer it started. A call that fails at the operator is not charged.