Build on Sable.

Sable speaks the OpenAI API. If your code already targets /v1/chat/completions, you're 30 seconds away from running it on private infrastructure.

Start here

• Quickstart — make your first call
• Authentication — bearer tokens and sessions
• Chat completions — the main inference endpoint
• Embeddings — vector embeddings
• Streaming — Server-Sent Events
• Privacy tiers — standard, confidential, sovereign
• Models — what's in the catalog
• Webhooks — outbound event subscriptions
• Rate limits — token-bucket budgets
• Privacy contract — what we will and won't store