Smart routing
Ask for a capability like chat, code, vision or reasoning and every request lands on the best-value model that can do the job. Or name a model and get exactly that one.
Every model · one endpoint · free to start
Real-time inference at planet scale.
Deploy any model. Serve any volume. AI Inference puts an entire catalog of models behind one endpoint, routes every request to the right one, and hands you a receipt for every token.
- No card to sign up
- Free monthly allowance
- Cache hits never billed
- No seat charge
capability: "reasoning"
prompt: "Draft a launch plan…"Status● success
Tokens in / out412 / 1,086
Latencymeasured per call
Priceshown on every response
One endpoint
Every model in the catalog behind a single, stable API.
Every call receipted
Tokens, latency and price on each response, as it happens.
Cache hits free
Ask the same thing twice and the second answer costs nothing.
Free to start
A monthly free allowance and no card to sign up.
The platform
Everything production inference needs, in one place
Stop stitching together keys, dashboards and retry logic for every model you use.
Automatic fail-over
If a model stumbles, your request moves to the next one that fits, so your users get an answer instead of an error.
Instant response cache
Repeat prompts come back from your own cache in a blink, and cache hits are never billed.
Limits you control
Per-account rate limits and cache lifetimes you set yourself, so a runaway loop never becomes a runaway bill.
Live analytics
Daily tokens, spend and latency, month-to-date totals and request history with a receipt for every call.
Keys done right
API keys are shown once and stored only as a hash. Revoke any key instantly. Idempotency keys make retries safe.
How it works
From sign-up to production in three steps
- 1
Create your account
Sign up free in under a minute. Your environment spins up automatically while you watch.
- 2
Try it in the playground
Pick a model or a capability, run a prompt and see the answer, the tokens, the speed and the price.
- 3
Ship it
Create a key, point your app at one endpoint and go live. Scale from one request to millions without changing a line.
For developers
One request. Any model.
Send a prompt and the capability you need, or name the exact model. Add an idempotency key and a retry is never metered twice. One bearer key, one small JSON body, and you are live.
curl https://ai-inference-silk.vercel.app/api/ainf/inference \
-H "Authorization: Bearer $AI_INFERENCE_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: welcome-001" \
-d '{"prompt":"Write a haiku about speed","capability":"chat"}'Pricing
Start free. Grow on your terms.
Every model shows its per-token price before you call it. Plans add included volume each month; usage beyond it is metered and invoiced every 30 days.
Free
$0
Free monthly allowance, no card
- Full catalog and playground
- API keys and analytics
- Pay per token beyond the allowance
Launch
$99/mo
5.00M tokens included
- Included monthly volume
- Full console and analytics
- Email support
Scale
$499/mo
30.00M tokens included
- Higher included volume
- Priority routing
- Usage alerts and reporting
Enterprise
$2,499/mo
200.00M tokens included
- Committed capacity
- Named point of contact
- Custom terms and reporting
Solutions
Built for every industry
From AI agents to agriculture, see what teams in your field run on AI Inference.
AI Agents
Agents that think fast and never stall
Explore →SaaS
Ship the AI feature on this sprint
Explore →Marketing
More campaigns, less blank page
Explore →Finance
Close faster, explain everything
Explore →Healthcare
Less admin, more care
Explore →Retail & E-Commerce
A whole catalog, beautifully described
Explore →Education
More teaching, less prep
Explore →Real Estate
List it, lease it, close it faster
Explore →Manufacturing
Keep the line moving, and the paperwork too
Explore →Transportation & Logistics
Every exception handled before it escalates
Explore →Hospitality
Five-star replies, every single time
Explore →Government
Public service, in plain language
Explore →No surprises
Clarity on every request and every invoice
Know the price first
Every model shows what it costs you per token before you send a single request.
Pay for what you used
Metered per token. No seat charges, no minimums, and one clear usage invoice every 30 days.
Stay in charge of data
You decide what you send. Prompts are processed to serve your request and kept only as described in our privacy policy.
Leave any time
No contract on self-serve plans. Stop sending requests and you stop paying. It really is that simple.
Frequently asked questions
What is AI Inference?
One API and one console for a catalog of AI models. Send a prompt with the capability you need, or name a model, and get an answer with its tokens, latency and price attached.
Do I need to rewrite my code?
Barely. It is one HTTPS endpoint with a bearer key and a small JSON body: your prompt, plus the capability or model you want. Most teams move an existing integration over by changing the base URL, the key and a couple of field names.
How does billing work?
Usage is metered per token at the price shown for each model. Usage beyond your free allowance and plan is invoiced every 30 days and paid online in a couple of clicks. Cache hits are not billed.
Is there a free tier?
Yes. Every account gets a free monthly token allowance, and you do not need a card to sign up. The remaining balance is always visible in your dashboard.
What happens if a model is busy or down?
Your request fails over to the next model that matches the capability you asked for, so it completes instead of erroring.
What if an invoice is not paid?
You get a few days after each invoice is issued. If a balance goes past due, usage pauses until it is paid in full, and everything resumes the moment it is.
Your next model is one request away.
Free to start, live in minutes, and a bill that is exactly what you used.