Plans and limits

OpenLLM has three plans: Free, a one-time Trial of Pro, and Pro. For the current Pro price, see www.openllm.sh/pricing.

Plans

• Devices: only running daemons count. See Device limits.
• Included models: allowances are counted in fresh-input tokens, with output and cached input weighted per model. See Included models.
• Tool calls: see Marketplace.

All monthly counters reset on the 1st of the month at 00:00 UTC.

Monthly tokens

On Free, requests to your own providers share a budget of 50M non-cached tokens per month. Trial and Pro have no monthly cap.

Each request counts:

max(0, input tokens − cached input tokens) + output tokens

Cached prompt tokens don't count, so a prompt cache hit costs you nothing against the budget. Tokens written to a prompt cache do count, because they are fresh input.

What counts:

• Successful calls to /v1/chat/completions, /v1/messages, and /v1/responses served by your own providers, including requests your devices serve through subscriptions and local models.

What doesn't count:

• Included openllm/ models. They have their own allowances, listed above.
• Embeddings, audio, images, video, and search.
• Failed requests.

At 80% of the budget, the app warns you with how many tokens are left and the date it resets.

When the budget runs out, OpenLLM refuses inference requests until the 1st of next month (UTC). This includes included models and the endpoints that don't count toward the budget. Requests get 402 with the type free_tier_exhausted:

{
  "error": {
    "message": "OpenLLM: payment required — free tier exhausted",
    "type": "free_tier_exhausted",
    "tokens_used": 50012345,
    "tokens_limit": 50000000,
    "period": "2026-10"
  }
}

To keep going before the reset, upgrade to Pro, which removes the cap.

The budget is a soft cap. Usage is checked against a total that can lag by a few seconds, so requests in flight when you reach the limit can take you slightly past it.

Trial

The trial gives you Pro for 3 days.

• Card required. You start the trial at Stripe Checkout and enter a card there. Nothing is charged during the trial.
• One per customer. The trial is offered only if you have never had a subscription. If you had Pro before, Resume Pro charges you right away.
• It turns into Pro. At the end of the trial, the subscription renews as Pro at the current price unless you cancel. If you cancel during the trial, it ends at the trial's end with no charge.

To start, open Settings → Billing → Plan and select Start free trial. If you have a promo code, enter it under Have a promo code? instead and select Continue with code. The code's terms are shown at checkout, and a code that makes Pro free replaces the trial.

If OpenLLM sent you an invite by email, enter the code from that email in the same field. Invite access doesn't need a card, and it ends on its own unless you add one.

Billing

Billing runs through Stripe. Everything is in Settings → Billing:

• Plan shows your plan and its status. Manage billing (or Manage / cancel during a trial) opens the Stripe customer portal, where you change your payment method or cancel.
• Invoices lists your past invoices.

When you cancel, Pro stays active until the end of the current billing period. The plan then shows Active until that date, and you can select Resume to keep it. After it ends you're back on Free.

Subscriptions renew automatically each period until you cancel. Under the Terms of Service, fees are non-refundable except where the law requires or a separate refund policy says otherwise. OpenLLM may grant refunds or credits case by case. To ask, email [email protected].

Related

• Included models
• Device limits
• Marketplace
• Errors

Last updated on