Every model currently routable through Northern Inference. Copy the Routing key into your OpenAI-compatible client (Cursor, Aider, Continue, Zed, LibreChat, LangChain, …) and POST to /v1/chat/completions. Anthropic-native clients work too: same base URL, POST to /v1/messages with native model IDs (claude-sonnet-4-5-20250929). That's the exact string you put in the model: field.
Managed Canadian cloud routes. Use when Canadian data residency is required.
Provider-default or global routes. Use for broader model access outside Canadian residency.
Northern Inference publishes a per-route data-use posture for every model in the catalog. Each claim is sourced from the provider's published data-governance documentation and dated. Click the Data use button on any model card to see that route's specific posture. The posture system has four dimensions:
Whether the provider contractually commits to not using your prompt and completion data to train or fine-tune AI models. Verified against each provider's data-governance terms.
Whether the provider logs or samples prompt and completion content after the request completes. Includes the provider's abuse-monitoring posture, which may involve human review of a content sample.
How long the provider retains any logged content. Zero means content is not retained after the request. Routes with an approved modified abuse monitoring agreement may have a shorter retention window.
Applies to BYOK routes: the customer's own provider agreement governs data use. Northern Inference makes no data-use warranty for BYOK routes. The claim row will indicate the governing agreement.
Claims show "Not yet verified for this route" when Northern Inference has not yet sourced a provider artifact for that dimension. A source link and as-of date are shown for each verified claim. Postures are updated when provider documentation changes. For the full data-use policy, see our Privacy Policy.