API

Models

The catalog is public. Agents can discover models and prices before they spend anything.

List models

curl

curl https://arcrouter.co/models

response

{
  "data": [
    {
      "id": "gpt-oss-120b",
      "provider": "arcrouter",
      "type": "chat",
      "modality": "text",
      "context_length": 131072,
      "min_output_tokens": 512,
      "input_usdc_per_mtok": 0.38,
      "output_usdc_per_mtok": 1.5,
      "released_at": "2025-08-05"
    }
  ]
}

GET /v1/models returns the same entries in the OpenAI list shape, so SDK model pickers work.

Fields

type: chat, embeddings or decisions, which decides the endpoint. Decision models are covered in Decisions (Jev).
input_usdc_per_mtok / output_usdc_per_mtok: the posted price per 1M tokens. This is what you are charged.
context_length: the window. It also bounds the pay-per-call ceiling.
min_output_tokens: reasoning models think before they answer. A max_tokens below this is rejected up front instead of billed for an empty reply.
private: true for models served inside a hardware enclave. See below.

Private models

Models named private/* run inside a GPU enclave (Intel TDX with an NVIDIA confidential GPU), served by Phala. The machine's operator cannot read memory inside the enclave, and the hardware signs a report of exactly what code is running. On our side, private calls are never cached and nothing about them is logged except token counts and the charge. They are called and paid for exactly like any other model.

Every private response carries proof it ran in the enclave:

X-ArcRouter-Receipt: the receipt id. The receipt's chat_id equals the completion id.
X-ArcRouter-Receipt-URL: fetch the signed receipt. It is public and free, and holds hashes and timestamps only, never content.
X-ArcRouter-Attestation: the enclave's attestation report, served directly by Phala. Its receipt_signing_keys are the keys the receipt must be signed with, so you never have to trust us to check it.

What this protects against: the machine's operator, and any record of your prompts being kept. What it does not: ArcRouter handles your request in memory on its way to the enclave, the same as any gateway.

Embeddings

curl

curl https://arcrouter.co/v1/embeddings \
  -H "Authorization: Bearer $ARCROUTER_KEY" \
  -H "content-type: application/json" \
  -d '{"model":"text-embedding-3-small","input":["first text","second text"]}'

Browse the live list with prices on the Models page.

NextDecisions (Jev) →