Advanced

Custom models

Publish your own model identifiers to Forge. Use it to alias upstream models, expose fine-tunes, or front a self-hosted endpoint with the same OpenAI schema.

Use cases

Pin a stable internal name (acme/chat-prod) that maps to whichever upstream model wins your evals this month.
Expose a private fine-tune to your team without leaking the upstream ID.
Set custom per-token prices for showback / chargeback across business units.

Create one

Go to Models → Add in the console. Fill in name, vendor, protocol, context window, and prices. Attach features (vision, tools, json mode) so the playground and SDK metadata respond correctly.

Call it like any other model

typescripttypescript

await ai.chat.completions.create({
  model: "acme/chat-prod",
  messages: [{ role: "user", content: "hi" }],
});

Authorize who can see it

Custom models are private by default. Use Authorize to grant access to specific users or workspaces, or publish to make them callable by any key in your tenant.