SEC.01 — Pricing
Per-token rates depend on the model tier, your region, and whether you're on shared or dedicated capacity. We quote against the shape you actually run, lock it for the term, and report the applied rate on every response.
SEC.02 — PLANS
Every plan reaches the same catalog through the same schema. What changes is the rate, the capacity you can reserve, and how much paperwork your security team gets.
PLAN 01 — Pay-as-you-go
Free to start / no base fee
Per-token billing across the full catalog. No commitment, no minimums — start with a trial key, scale when you're ready.
| Included |
|---|
| All catalog models |
| Shared inference pools |
| Streaming, tools, vision, JSON mode |
| Email support, business hours |
| 10K req/min default ceiling |
PLAN 02 — Most teams
Volume commitment / monthly term
Volume commitment unlocks priority routing, lower per-token rates, and reserved capacity headroom for peak traffic.
| Included |
|---|
| Locked per-token rates for the term |
| Priority routing & failover |
| Per-tenant rate budgets |
| SLO with credit-back guarantee |
| Shared Slack channel |
PLAN 03 — Reserved capacity
Custom / monthly term
Reserved replicas, BYO-provider, data-residency, zero-retention routes, and the paperwork your security team requires.
| Included |
|---|
| Dedicated GPU replicas |
| Custom regions & data residency |
| Zero-retention inference routes |
| BYO-provider key support |
| SOC 2 type II artifacts on request |
SEC.03 — PER-MODEL RATES
The rate you pay is set by your contract and the route your call takes — region, capacity type, and load all move it. Whatever it lands at, it comes back in the response headers of every request, so nothing has to be reconstructed later.
FIG.01 — METERING PATH
Per-model rate bands → MODEL CATALOG · DWG models.html
Open model catalogCommitted customers run on a single flat rate for the whole term
The exact rate for your term is set in your order form
Dedicated capacity is priced per replica-month and quoted separately
SEC.04 — ADD-ONS
Most workloads only need the per-token meter. These line items show up for teams who need more control over where their inference runs.
| Add-on | What it covers | How it's priced |
|---|---|---|
| Dedicated replica | Reserved GPU capacity for a single model, in a region of your choice. | Quoted per deploymentper replica · month |
| Zero-retention route | Prompt and completion bytes never leave volatile memory. Audit log only. | Uplift on token ratesset in your order form |
| Data-residency lock | Requests pinned to a specific region with hard failover off. | Small uplift on token ratesvaries by region |
| Premium support | 24/7 paging, named contact, quarterly architecture review. | Flat monthly feescoped to your team size |
| BYO-provider routing | Use your own provider key under our schema, routing, and observability layer. | Metered per requestvolume-tiered |
Add-ons are quoted with your token rates · what applies to your account is set in your order form
SEC.05 — PRICING FAQ
Because the rate genuinely varies. A request routed to a US-West shared pool at off-peak isn't priced the same as a Tokyo dedicated replica under peak load, and commitment moves it again. Publishing one number would mean either undercharging us out of business or overcharging customers who don't need the premium path. So we quote against your workload — and every response carries the rate that was actually applied, in its X-IDC-Rate-* headers, matching your invoice line for line.
You commit to a monthly token volume (or a spend floor) for a term of 3, 6, or 12 months. In exchange, your per-token rate is locked at a single number for the duration, no matter how the underlying route moves. Unused commitment doesn't roll over by default; we can quote a roll-over clause if it matters to your forecasting.
Per token, as reported by the upstream provider, with no rounding-up. Cached prompt tokens are billed at a discounted cached rate that varies by model. Tool-use and structured-output overhead is included in the token count; we don't charge a separate "agent surcharge."
Yes — every new account starts with a trial credit large enough to validate a real workload, not just hello-world. Email sales@idclinktech.com with a short description of what you'd test and we'll size the credit accordingly.
Credit card, ACH/wire, and invoice (net-30) for committed customers. We can support most procurement systems and master service agreements; talk to sales for the paperwork.
SEC.06 — QUOTATION
Send us your monthly token estimate, model mix, and region. We'll come back with a locked-rate offer for that exact shape — usually within one business day.