Overpaying for easy work
Sending every request to your heaviest model is the most expensive habit in AI.
Routing
The right model, every request
Tell us what the request needs, chat, code, vision or reasoning, and the router picks the lowest-priced enabled model that can do it. If that model stumbles, another one catches the request.
Lowest-priced fit
chosen per request
Automatic fail-over
busy models do not fail your call
Zero code changes
as the catalog evolves
The problem
What is slowing you down today
Hard-coded model names
Every change in the model landscape becomes a pull request.
One busy model, one failed request
Without fail-over, a moment of congestion becomes an error your user sees.
Capabilities
Everything you need, nothing you don't
Route by capability
Send capability in the body. Leave it out and chat is assumed.
Lowest price first
Eligible models are tried cheapest first, so routine work stays affordable.
Smart fail-over
Timeouts, rate limits and server errors move the request to the next eligible model. Errors that would fail everywhere, like a malformed request, return immediately.
Pin when you need to
Pass a model_id to use exactly that model, with no substitution.
Account default
Set a default model in settings for requests that name none.
See what served you
Every response names the model that answered, with its tokens and price.
Use cases
What teams run on it
01
Mixed workloads
Cheap models for classification, reasoning models for analysis.
02
Resilient features
User-facing calls that ride out a busy model.
03
Cost control
New lower-priced models are picked up without a deploy.
04
Strict workflows
Pinned models where output consistency matters most.
How it works
From sign-up to production
- 1
Send a capability
Add capability to your request body, for example code or vision.
- 2
Let the router choose
The lowest-priced enabled model with that capability serves the request.
- 3
Check the response
model_id and model_name tell you which model answered.
Control
Built so you stay in charge
Keys you control
API keys are shown once at creation and stored only as a hash. Revoke any key instantly from Settings.
Spend you can see
Every request carries its own receipt: tokens in, tokens out, latency and price. Nothing is billed that you cannot trace.
Limits you set
Per-account rate limits and a free monthly allowance keep experiments safe before they become production.
Data you decide on
You choose exactly what you send. Repeat prompts can be served from your cache, and cache hits are never billed.
For developers
One endpoint. Every model.
Send a prompt and a capability, or name a model. One bearer key, one small JSON body, and a full receipt on every response.
curl https://YOUR-DOMAIN/api/ainf/inference \
-H "Authorization: Bearer $AI_INFERENCE_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: order-1042" \
-d '{"prompt":"Summarise this ticket in one line","capability":"chat"}'Questions, answered
What capabilities can I request?
Whatever the enabled catalog offers, such as chat, code, vision and reasoning. The model catalog lists each model's capabilities.
Does fail-over cost extra?
You pay for the call that succeeds, at the price of the model that served it. Failed attempts are not billed.
Can I turn routing off?
Yes. Pass a model_id and the router uses that model only.
What if no model can serve the request?
You get a clear error. If no enabled model has the capability, the response says so.
Stop picking models by hand
Send a capability and let the router do the rest.