momor.ai
Open Search

For inference providers

Serve your models to every Gateway customer.

Customers ask for a model. Gateway sends each request to a provider that serves it, weighed on price, speed and reliability. Move the sliders to see how your share of the traffic changes.

Sign in to applyRead the requirements
llama-4-maverick · share of requestssample
  • Your endpoint · 27%$0.58 · 0.5 s · 99.90%
  • Groq · 35%$0.60 · 0.3 s · 99.95%
  • Fireworks · 22%$0.60 · 0.6 s · 99.90%
  • Together · 17%$0.62 · 0.7 s · 99.80%
Your requests per day312,376
Estimated earnings per month$4,892

Becoming a provider

Apply, pass the tests, start receiving traffic.

You need an OpenAI-compatible endpoint, published prices per million tokens, and the precision and context length you serve each model at.

  1. Apply
    • Company and contacts
    • Models and prices
    • Endpoint and credential
    • Data policy
    about 15 minutes
  2. Tests
    • Correct answers
    • Tool calls
    • Structured output
    • Speed under load
    • Error rate
    about 3 days
  3. Approval and traffic
    • Listed in the catalog
    • Weighed in routing
    • Health checks continue
    when tests pass

Testing

Every model is tested before it gets traffic.

Momor sends test requests for correctness, speed and errors before routing customer traffic, and keeps checking after you go live.

Test report · llama-4-maverickOct 18 · 1,200 requests
  • Answers match the reference400 prompts with known answers99.2%Passed
  • Tool calls200 requests with tools100%Passed
  • Structured outputJSON schema requests96.0%Failed
  • First token, medianat 50 requests per second0.48 sPassed
  • Error rateover 1,200 requests0.1%Passed
  • Context lengthup to 1M tokens as listed1MPassed

One check failed. Fix it and rerun the tests from Settings › Provider; traffic starts once everything passes.

Routing around

If your endpoint fails, traffic moves, and you see why.

When your endpoint errors or slows down, Gateway routes around you and records it: the model, the error and how long it lasted.

Capture · Settings › Provider › Health
The Health page in Settings › Provider, with each time Gateway routed around the endpoint: the model, the error and how long it lasted.

Payouts

Paid for the traffic you serve.

A monthly statement per model. Payouts go through Stripe, which handles identity, bank and tax details.

Statement · September 2026Paid Oct 15 via Stripe
ModelRequestsTokensEarned
llama-4-maverick9.4M8.1B$22,140.60
qwen3-coder2.1M2.6B$7,812.40
gemma-4-26b0.8M0.6B$1,465.20
Paid out$31,418.20

Data policy

Your data policy is shown to customers.

Whether you log prompts, whether you train on them, how long you keep them and where they’re processed.

What customers seeon every model you serve
Your companyServes 3 models on Momor Gateway
Logs prompts
No
Trains on prompts
Never
Keeps data
0 days
Processed in
US, EU

Customers can exclude providers whose policy doesn’t meet theirs. Changes are shown with notice.

In Settings

Models, health and earnings, in your Settings.

Pause a model, change a price with notice, replace your credential, and see what you’ve earned.

Capture · Settings › Provider
The Overview page in the provider organization’s Settings: the models served, their health and what they have earned.

List your first model.

Apply with your endpoint, models and prices. We review applications within 5 business days.

Apply to list your modelsTalk to us
Or read about Momor Gateway