Usage
Models
Real model names, EU location, context length, capabilities — the aliases stay stable when we update the model behind them.
Aliases instead of version numbers
You address models through aliases — futureway-fast, futureway-smart, futureway-bigthink and futureway-embed. They stand for a performance class, not a model version: moving to a better model of the same class does not change your code.
Chat models
| Alias | Model | Class | Context (tokens) | Capabilities | Location |
|---|---|---|---|---|---|
futureway-fast | Mistral Small 3.2 24B Instruct | fast | 128,000 | images · tools | Paris, FR 🇪🇺 |
futureway-smart | Mistral Medium 3.5 128B | standard | 180,000 | images · tools | Paris, FR 🇪🇺 |
futureway-bigthink | Qwen3.5 397B A17B | thorough | 250,000 | images · tools | Paris, FR 🇪🇺 |
Embedding models
| Alias | Model | Location |
|---|---|---|
futureway-embed | Qwen3 Embedding 8B (1024 dim) | Paris, FR 🇪🇺 |
The vector dimension is in the response; it does not change without notice in the changelog.
Model list via API
The current list comes from GET /v1/models — exactly the aliases your key may address.