Models & sovereignty

Discover models, pricing, and data-residency tiers.

Models are identified in provider/model form, e.g. mistral/mistral-medium-3.5. List every active model with GET /v1/models. The list is the same for every key; your privacy settings and guardrails are applied when you call a model, not here:

json

{
  "object": "list",
  "data": [
    {
      "id": "mistral/mistral-medium-3.5",
      "object": "model",
      "owned_by": "mistral",
      "kind": "chat",
      "context_length": 262144,
      "sovereignty": "strict",
      "zdr": "partial",
      "no_training": true,
      "forbidden_segments": [],
      "pricing": {
        "input_micro_eur_per_mtok": 1250000,
        "output_micro_eur_per_mtok": 6400000
      },
      "supported_parameters": ["max_tokens", "temperature", "top_p", "stop", "stream", "tools", "tool_choice"]
    }
  ]
}

Fields

Field	Meaning
`kind`	`chat`, `embedding`, `transcription` or `speech`. Decides which endpoint accepts the model.
`context_length`	Maximum tokens per request, where the provider publishes it.
`sovereignty`	`strict` or `pragmatic`, see below.
`zdr`	Zero data retention at the provider: `default` (the provider keeps nothing), `partial` (limited retention, for example abuse logs) or `none` (no zero-retention commitment established or assessed; the provider may keep prompts).
`no_training`	`true` if the provider contractually excludes training on your data.
`forbidden_segments`	Regulated usage segments the provider excludes by contract (for example `health`). If your organization is registered for one of them, requests to this model are refused with `403 segment_not_permitted`.
`pricing`	Prices in micro-EUR: per million input and output tokens for chat and embeddings, `audio_micro_eur_per_hour` for transcription, `character_micro_eur_per_million` for speech.
`supported_parameters`	The request parameters this model honours. Test for `tools` before sending tools and for `reasoning` before asking for reasoning.
`reasoning`	Present only on models that think. See Reasoning.

Sovereignty tiers

Each model declares a sovereignty tier so you can choose how strict your data residency requirements are:

Tier	Meaning
`strict`	An EU-owned provider that Kontinent rates as sovereign.
`pragmatic`	Served from the EU, but by a provider with a non-EU parent or one that runs on third-party cloud infrastructure.

The tier is returned on every model, so you can require strict where the use case demands an EU-owned provider and allow pragmatic elsewhere. See Maximum privacy.

Model kinds

Kind	Endpoint
`chat`	`POST /v1/chat/completions`, `POST /v1/responses`
`embedding`	`POST /v1/embeddings`
`transcription`	`POST /v1/audio/transcriptions`
`speech`	`POST /v1/audio/speech`

Sending a model to an endpoint of another kind is rejected; an unknown or disabled id returns 404 model_not_found.

Pricing and the money unit

Money everywhere in Kontinent is integer micro-EUR, so there is no floating-point rounding in billing:

1 EUR = 1,000,000 µEUR

Prices are quoted per million tokens, split into input and output rates, and returned on every model. In the example above, input costs 1,250,000 µEUR/Mtok, which is €1.25 per million input tokens, and output costs €6.40 per million output tokens. You pay only for the tokens a request actually uses.

Two rules lower or raise that price and are not part of /v1/models:

Cached input. When a provider serves part of your prompt from its prompt cache and reports it, those tokens bill at the model's lower cached rate. The count appears as cached_tokens in the logs.
Long context. A few models have a higher rate above a prompt-length threshold. Once a prompt reaches it, the whole request bills at that rate, as the provider meters it.

See Rate limits & credits for how spend is metered.

API reference Price & performance Maximum privacy

Fields

Sovereignty tiers

Model kinds

Pricing and the money unit

Related