Open models. Operated by Rackspace.

Your AI. Your rules.
Our responsibility.

Run open models in production without becoming the AI operator. Rackspace deploys, scales and runs them on Rackspace-operated infrastructure. Your models, data and boundaries stay yours.

Start building free →Talk to an AI architect →
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://<your rackAI gateway>/v1/tenants/<tenant>/openai/v1",
    api_key=os.environ["RACKAI_API_KEY"],
)

resp = client.chat.completions.create(
    model="<deployment>",
    messages=[{"role": "user", "content": "Summarize this SOW"}],
)  # OpenAI-compatible. Swap one line.

Why production AI stalls

AI is easy to try. Hard to run.

A working model is only the start. A dependable service needs performance, security, cost control, safe change, and someone accountable when it breaks. rackAI puts all of it under one operator.

You define the boundaries. rackAI operates within them.

  1. You define:

    • objectives and priorities
    • policies and boundaries
    • your data and models

  2. rackAI operates:

    1. deploy
    2. optimise
    3. govern
    4. assure

  3. You can verify:

    • performance
    • activity
    • cost
    • adherence to your policies

How rackAI is designed to work: one agreement drives placement, serving, monitoring, cost and evidence.

What changes for you

From pilot to accountable production.

Move to production faster.

Use the OpenAI-compatible tools you already know. Rackspace handles deployment, scaling and day-to-day operations.

API compatibility

Keep control.

Your models, data, access and deployment boundaries remain under your authority, with isolation and policies built into how workloads run.

Security and assurance

Operate with confidence.

Rackspace owns platform operations. You get visibility into performance, usage, changes, and the evidence behind them.

Models and deployment

Improve continuously.

Fine-tune models, optimise performance and resource use, and evolve workloads as your needs change—with human oversight where it matters.

Your data and fine-tuningThe rackAI FDE

The rackAI commitment

We commit to the outcome.

99.9%
Monthly availabilityPer eligible production deployment, with service credits.
< 1 hour
Service restorationFollowing an eligible platform-side serving failure.
< 10 min
Time to first inferenceFrom a provisioned, ready environment.
Monthly
Operational evidencePerformance, usage, changes and accountability.

One accountable operator. Clear commitments. Evidence you can inspect.

Talk to an AI architect →

Illustrative presentation of proposed commitments. Eligibility, measurement, remedies and reporting are subject to agreed service terms.

Get started

From first conversation to your first call.

  1. Get access

    Talk to us for an Organization, or sign in to yours and create an API key.

    API keys and authentication

  2. Deploy a model

    Choose an open model and deploy it. Try it in the Playground first.

    The model catalogue

  3. Switch your code

    Change your OpenAI SDK's base URL and key, and ship.

    Using your existing SDK

Ready when you are

Your first million tokens are on us.

1,000,000 trial tokens for your Organization, shared by every Tenant. Test your integration on your own deployment before you commit to production.

Trial details
Claim your 1M trial tokens →