Plans and pricing
NuRoute charges for two things, separately:
- Your plan pays for the gateway: routing, governance, dashboards and request traces. Free, Starter, Team or Enterprise.
- Credits pay for models on our keys. This includes Artemis Task (
model: "auto") and any named model you run on NuRoute's provider keys. With your own provider keys, the provider bills you and no credits are used.
A plan never includes model usage, and credits never pay for a plan.
Gateway plans
| Free | Starter | Team | Enterprise | |
|---|---|---|---|---|
| Price | $0 | $49 / month | $299 / month | Custom |
| Annual | – | $490 / year | $2,990 / year | Contract |
| Requests included | 10,000 / month | 250,000 / month | 2,000,000 / month | Agreed volume |
| Extra requests | Paused at the limit | $0.15 per 1,000 | $0.10 per 1,000 | Contract |
| Projects | 1 | 3 | Unlimited | Unlimited |
| Members | Unlimited | Unlimited | Unlimited | Unlimited |
| Routing, Artemis, OpenAI-compatible API | ✓ | ✓ | ✓ | ✓ |
| Your own provider keys | ✓ | ✓ | ✓ | ✓ |
| Request detail kept | 7 days | 30 days | 90 days | Custom |
| PII redaction | – | ✓ | ✓ | ✓ |
| Budget guards (cost caps, project budgets) | – | ✓ | ✓ | ✓ |
| Org policies (provider rules, limits) | – | – | ✓ | ✓ |
| Data residency rules | – | – | ✓ | ✓ |
| Single sign-on | – | – | – | ✓ |
| Admin audit log | – | – | – | ✓ |
| Self-hosted deployment | – | – | – | ✓ |
| SLA | – | – | – | ✓ |
| Support | Community | Shared Slack | Dedicated, SLA |
All prices are in US dollars, plus VAT where applicable. Annual billing gives you two months free.
Model usage: prepaid credits
| What runs | Price | Paid from |
|---|---|---|
Artemis Task (model: "auto") | $0.75 per 1M input tokens, $3.20 per 1M output tokens, $0.03 per 1M cached input tokens | Credits |
| A named model on NuRoute's keys | The provider's list price, with no markup | Credits |
| A named model on your own provider key | Nothing from NuRoute; the provider bills you | — |
- Artemis Task always runs on NuRoute's provider keys, whichever model it picks. Reasoning ("thinking") tokens count as output. Cached input is charged at the cached rate only where the serving provider actually gives a discount for it; otherwise it is charged as normal input.
- Try it free: new organizations get $2 of credits once the email address is verified (straight away with Google sign-in).
- Credits are prepaid in US dollars by card, never expire, and unused purchased credits are refunded when you close your account. Each top-up gets a Stripe invoice.
- When credits run out, Artemis Task and requests on NuRoute's keys return
402 insufficient_creditsuntil you top up. A request that names a model you have your own key for still runs, on your key. Requests already running finish.
What counts as a request
One request is one call to an inference endpoint that NuRoute accepts and routes, whichever model answers it.
| Case | Counts as |
|---|---|
| A successful request, on any model or key | 1 |
| A streamed request | 1 |
| A request retried on another provider after an error | 1 in total |
| A request every provider failed after it was routed | 1 |
| A request refused before routing (bad key, invalid request, policy block, no credits, plan limit, rate limit) | 0 |
| Management calls (models, keys, usage) | 0 |
Free counts per calendar month (UTC); paid plans count per billing period.
Going over your plan
- Free pauses at its monthly limit: further requests return
429 plan_request_limit_reacheduntil the next month, or until you upgrade. - Starter and Team keep running. Each started block of 1,000 extra requests is charged from your credits (Starter $0.15, Team $0.10 per 1,000).
- Hard cap: turn on Stop at my plan's limit on the Billing page to pause instead of paying overage.
- Extra requests also pause when you have no credits to pay for them.
- Owners get an email, and the dashboard shows a banner, as you near the limit and when you reach it.
Changing plans
| Change | When it applies | Money |
|---|---|---|
| Upgrade | Immediately | You pay the difference for the rest of the period |
| Downgrade or cancel | At the end of the paid period | No refund for the current period |
| Monthly → annual | Immediately | The unused monthly part is credited |
| Annual → monthly | At the annual renewal | — |
Owners and admins change the plan from the Billing page. Team can also pay by invoice (bank transfer or SEPA, 14 days): ask us.
Payments and downgrades
- A failed card payment is retried for about two weeks, and the plan stays active meanwhile.
- An unpaid invoice gets 7 days of grace after its due date.
- After that, the organization moves to Free. Nothing is deleted, and your credits stay. Paying restores the plan.
When your plan no longer includes a feature you set up:
- PII redaction, policies, data residency rules and budgets keep being applied. You can remove them, but not change or add them.
- Existing projects keep working; you can't create new ones over the limit.
- Single sign-on keeps working for 14 days, then members sign in with password or Google.
Data retention
| Data | Kept |
|---|---|
| Request detail: the stored prompt and response, and the request in the Request Explorer | 7 / 30 / 90 days by plan; custom on Enterprise |
| Usage numbers (requests, tokens, cost) | 90 days in the dashboard; 12 months for reporting |
| Credit and payment records | 10 years, as German tax law requires |
When your plan's window gets shorter, older detail is removed only after 14 days.
Enterprise
Custom request volume, retention and price, single sign-on, the admin audit log, a self-hosted deployment and an SLA. Artemis Task is charged at the same price per token, and credits can be bought by invoice. Self-hosted installs run the full gateway; for Artemis on self-hosted, talk to us. Book a call.