Skip to sign-up
AtriumLLM
Pricing

Pay for the foundation. Keep your keys.

Start with a dedicated instance that sleeps when idle, add your people with Teams, and move into your own cloud when you’re ready. Model usage is never marked up.

Hosted Starter

For growing companies that want a model layer on day one.

$149/month
250K requests included, then $10 per additional 100K. Sleeps when idle.
  • Your own dedicated, isolated instance
  • Every major model provider, your keys or ours
  • Routing, fallbacks and streaming
  • Traces and outcome tracking
  • Memory sovereignty: your data, fully exportable
Join early access

Hosted Growth

For production and real-time workloads.

$499/month
1M requests included, then $10 per additional 100K.
  • Everything in Starter
  • Always-warm instance for real-time use
  • Region pinning
  • 90-day trace retention
Join early access

Teams

For companies giving employees AI with model choice and real controls.

$15/seat/month
Billed annually, or $18 monthly. 10-seat minimum. Model usage billed separately.
  • A dedicated hosted instance
  • Employee workspace with a policy-filtered model picker
  • SSO with Entra ID, Okta or Google, plus SCIM sync
  • Group policies and spend caps
  • Exportable audit log and admin console
Join early access

Enterprise

For organizations governing AI across business lines, in their own cloud.

$60K/year and up
Annual contract. One business line and 250 seats included; $10 per additional seat per month.
  • Everything in Teams
  • Payload catalog and policy by business line
  • ML endpoints: embeddings, predict, transcribe
  • Deployed in your cloud, with schema-aware data protection and an SLA
  • Full sovereignty or on-prem: +$40K per year

Design partners get 50% off the first year in exchange for a case study and a seat at the roadmap table.

See Enterprise

Model usage is never marked up. Bring your own keys and pay providers directly, or run on ours and pay the provider’s list price.

Compare plans

Capability Starter Growth Teams Enterprise
Dedicated, isolated instanceIncludedIncludedIncludedIncluded
Multi-provider routing and fallbacksIncludedIncludedIncludedIncluded
Bring your own model keysIncludedIncludedIncludedIncluded
Traces and outcome trackingIncludedIncludedIncludedIncluded
Always-warm instance for real-time use—Included—Included
Region pinning—Included—Included
Employee workspace and admin console——IncludedIncluded
SSO, SCIM and group policies——IncludedIncluded
Audit log export——IncludedIncluded
Payload catalog and business-line policy———Included
ML model endpoints———Included
Where it runsOur hostingOur hostingOur hostingYour cloud or on-prem
Full sovereignty tier———Add-on

Questions

Do I pay model providers separately?

With your own keys, yes: you pay providers directly under your own terms. On our keys, usage is billed at the provider’s list price with no markup.

What counts as a request?

Each call to an AtriumLLM endpoint, such as a chat completion, an embedding or a transcription.

What happens when a Starter instance is idle?

It sleeps, so a quiet month costs only the plan fee. Growth keeps your instance warm for real-time use.

Can I move from Hosted to my own cloud later?

Yes. Every deployment uses the same template, and your traces, artifacts and outcomes are fully exportable.

Is my data used for training?

No, on every plan.

Get started with early access.