Skip to content

Dedicated deployments: Your model, tuned to the GPU kernel

Bring your own weights or any Hugging Face model to a dedicated endpoint, billed per GPU hour.

We reply by email.

Backed by Y Combinator,our optimization used by

NVIDIAHugging FaceSamsungAMDModalReddit

What you get

One endpoint, tuned for your model

Your endpoint

Capacity that serves your workspace only.

Tuning

It serves first, then we tune it, with no downtime.

Price unit

The GPU hour, paid three hours at a time from your balance.

Start a deploymentSee pricing

Serving stack

Five layers, tuned for your model

serving stack

The lab publishes its kernel search. AutoKernel paper (opens in a new tab), AutoKernel repository (opens in a new tab), AutoMegaKernel paper (opens in a new tab)

Start a deployment

How it goes

From the request to your endpoint

Start a deployment

Request

Tell us the model and the traffic

What happens next

We reply by email. The reply holds a plan, its price and its date, or the reason it cannot be served yet.

Send no token, key or weights in this form.

All fields are required.

A Hugging Face repository id such as org/name, a listed model, or your own words.

Requests a minute, tokens a request or hours a day, in your own words.

By sending, you confirm your right to run this model commercially under our Terms of Service.

Get started

$1 free for a new account.

Get an API key
rightnow
RightNow open models
Billed per token.
import OpenAI from "openai";const client = new OpenAI({baseURL: "https://api.rightnowai.co/v1",   apiKey: process.env.RIGHTNOW_API_KEY,      });const response = await client.chat.completions.create({model: "MODEL_ID",                         messages: [{ role: "user", content: "Hello" }],max_tokens: 16384,});