Prepaid credits
Top up in euros and pay only for the input and output tokens you use.
Get startedSimple, transparent pricing
Pay for the model that answers each request, or reserve dedicated capacity for predictable workloads.
Model rates
One API, eight models.
Prices below are in euros per 1 million tokens. A 5% CYANIS orchestration fee is added to the listed rate.
| Model | Input | Output |
|---|---|---|
| GPT-6 Luna | €0.087696 | €0.438481 |
| Ministral 3 14B | €0.175392 | €0.175392 |
| Mistral Medium 3.5 | €1.315443 | €6.577217 |
| Z.ai GLM 5.3 | €1.227747 | €3.858634 |
| Claude Haiku 5.5 | €0.089469 | €0.447347 |
| Claude Sonnet 5.5 | €1.789389 | €8.946945 |
| Claude Opus 5.5 | €3.578778 | €17.893889 |
| Claude Fable 5.1 | €8.946945 | €44.734723 |
Pricing FAQ
How are model rates calculated?
Rates are the list price per 1 million input or output tokens, converted to euros from provider pricing. A 5% orchestration fee is added.
Which models are available?
The catalogue includes models from Mistral, OpenAI, Z.AI and Anthropic. Availability is exposed through the API model endpoint for your account.
Are all models available in Sovereign mode?
No. Sovereign routing uses eligible EU-resident models. OpenAI, Z.AI and Anthropic models are available through standard routing but are not sovereign-eligible.
Is there a free tier or trial?
No. CYANIS Cloud uses prepaid credits for pay-as-you-go inference, or a capacity agreement for larger workloads.
How does Batch pricing work?
Asynchronous processing is available at up to 50% below real-time pricing.
What comes with Reserved capacity?
Reserved capacity includes a monthly committed volume, guaranteed capacity and a contractual SLA.
Does Claude Haiku have a prompt limit?
Yes. Claude Haiku 5.5 requests are limited to prompts of about 100,000 tokens under the listed flat rate.