Become a provider
Sell your inference capacity on BatchRouter. Who can join, what you need before you start, and how pricing and payouts work.
BatchRouter routes customers' AI work to provider lanes. Most of it is batch work: whole batches with a deadline, which you can hold and run in spare capacity. Customers can also send single requests that want an answer right away. As a provider, you serve that work from your own endpoint, set your own price per 1 million tokens, and get paid monthly in USD for the work you complete. Setup is self-serve and every check is automatic.
Who can join
Providers must be a registered company, a data center, or a professional edge network. Individual machines and hobby nodes aren't eligible for provider review or payouts. You confirm this when you register.
What you need
- A public
https://endpoint, plus a dedicated API key that BatchRouter will send with every request. Use a key you can revoke independently of your other customers. - One of two integrations. Both are described in full in the
Provider API:
- OpenAI Batch API (recommended). You host the OpenAI Batch API (files, batches and results) plus a few BatchRouter extensions for deadlines, pay, acceptance and partial results. This lane gets batch work, and it serves direct requests at its direct price.
- Chat completions only. An OpenAI-compatible server such as vLLM, TGI or SGLang that serves
/v1/chat/completions. This lane gets direct requests only, no batch work.
- The model IDs you serve and your price per 1 million tokens, in USD.
- Your data-handling policy: how long you keep prompts and outputs, whether you train on customer data, and the countries you run inference in.
- A bank account that can receive USD by international bank transfer.
How it works
-
Register at batchrouter.com/providers/apply with your company details, endpoint, integration type and data-handling answers.
-
Complete the setup checklist in the provider portal. Add your models and prices, test your endpoint, publish capacity, declare data handling and add payout details. See the onboarding checklist.
-
Go live. Go live re-runs every check, and routing to you starts automatically. There's no manual approval step. BatchRouter operators can switch a provider's routing off and on. New providers start with conservative routing limits.
-
Serve work. BatchRouter routes matching work to your lanes automatically: batches through your batch API, and direct requests through your chat completions endpoint. Your endpoint returns results to BatchRouter, which delivers them to the customer.
-
Get paid. Earnings are settled per calendar month and paid in USD by bank transfer, with a statement for every payout. See earnings and payouts.
Pricing and what you earn
You set a price per 1 million tokens for each model: input, output and, optionally, cached input. Embedding models take an input price only. That price is what you earn for completed work. BatchRouter adds its margin on top when it quotes customers.
On an OpenAI Batch API lane, that price is your batch price, and the lane also needs a direct price: what you earn for direct requests, and the reference your batch prices must stay below. On a chat-completions-only lane, the price you set is your direct price. See Prices.
Changing a price creates a new version of your offering. Quotes a customer already received keep the price that applied when they were quoted.
An account can have at most two providers in setup (registered but not yet live) at a time. Live providers don't count toward this limit.