Prices and reports

Set the internal price your organization's own keys are metered at, per model or as a discount, and read what your keys served, what fell back and what reached your credit.

Internal prices

Every model your keys serve has an internal price. It is the price Tokamak meters those requests at, so that analytics, Insights, budgets and your own chargeback show what the work cost your organization. Tokamak charges nothing for them.

Open Providers › Models & prices. You need org.models.manage, which owners and admins have. The table lists every model your keys serve:

Served by: which of your keys serves it.
Tokamak list: Tokamak's price per million input and output tokens, with the cache lines below.
Your internal price: what your keys' requests are metered at.
Source: List price, a discount, or custom rates.

The Models & prices tab: a note that internal prices only apply to requests your keys serve, tiles for models on your keys, custom prices, internal versus list and the last price change, and a table of models with Tokamak's list price, the internal price and its source — Six Anthropic models at a 20% contract discount; OpenAI models at list price except gpt-6-sol, edited below. Open full size ↗

Change a price

Choose Edit on a model. There are three sources:

Source	What it stores	Use it for
Tokamak list	No price of its own. It follows Tokamak's list, including future changes	The default
Discount %	The discount, applied to every line and stored as exact per-line prices	A committed-spend or volume discount
Custom rates	Your own input, output, cache-write and cache-read rates per million tokens	A negotiated rate card

The preview shows what the new price would have made the last seven days of this model's traffic cost. Charged by Tokamak stays $0. Budgets affected lists the budgets with a cost limit that count this model's key-served requests, since a new price changes how fast they fill.

The Edit internal price dialog for gpt-6-sol: discount 15% selected, Tokamak's list and the resulting price per line, and a preview of the last seven days at list and at the new price, charged by Tokamak $0.00 — A 15% discount on gpt-6-sol, with its effect on the last seven days and the budget it counts against. Open full size ↗

A price change is a new version:

It applies to new requests only. Every request keeps the price it was metered at, so a later change never rewrites history.
A discount is stored as exact prices, so a later change to Tokamak's list does not move it. To follow the list again, choose Tokamak list.
Bulk discount applies one discount to every model a provider's keys serve, such as your Anthropic contract.
Export price sheet downloads the table as a CSV file.

Reports

Everyone who reads usage sees the effect of your keys. Figures are over the window you choose.

Analytics › Credential source

In Analytics, switch to Credential source at organization scope. You need the organization admin or billing-view permission. It splits usage cost by who served it:

your keys, at your internal price;
Tokamak, at list price;
fallback, at list price.

The tiles show the usage cost, the amount charged by Tokamak, your keys' share of requests and the fallback rate.

Analytics with the Credential source view: tiles for usage cost, charged by Tokamak, via your keys, your-key share and fallback rate, and a daily chart of usage cost stacked by source — Test traffic: all 55 requests ran on the organization's keys and were charged $0.00. Open full size ↗

The Usage view of Analytics, and Insights, keep working as before. They count a key-served request at its internal price.

Billing & credits

Billing & credits shows what reached your credit over the last 30 days, by source. Requests your keys served are listed as Not charged. The Platform fee on your keys row shows the fee on those requests: 0%, so $0.00. Tokamak does not charge a fee on requests your keys serve.

Billing and credits: tiles for requests via your keys and a 0% platform fee, a note that your provider keys are billed by your providers, and a table of charges by source with Tokamak capacity, fallback from your keys, your keys, not charged, and the platform fee on your keys at $0.00 — Charges by source. In this test run, no request needed Tokamak capacity. Open full size ↗

Budgets & limits

By default a budget counts key-served requests at their internal price, so a team's budget still holds when its requests run on your key. Budgets & limits says so while BYOK is on.

Each budget chooses, under Requests served by your own provider keys:

Include your-key traffic (the default): key-served requests count against the budget and are refused once it is spent, like any other request.
Only Tokamak-charged spend: the budget caps what reaches your credit. Requests your keys serve are neither counted nor limited by it; fallback requests, which Tokamak charges, still are.

The budget form's choice for requests served by your own provider keys: include your-key traffic, selected, or only Tokamak-charged spend — The choice appears in the budget form while BYOK is on. Open full size ↗

A budget that counts your keys shows its spend split by source under Usage cost / budget:

An organization budget of $250.00 per calendar month named Engineering tooling budget: usage cost $0.3439 of $250.00, split into $0.00 Tokamak and $0.3439 your keys, 55 requests — Every request in this window ran on the organization's keys. Open full size ↗

A budget that leaves your keys out reads Tokamak-charged spend only.

Insights

At organization scope, Insights opens with your keys' share of metered requests and a Your-key share by member table: each member's requests, metered cost and the share their keys served. Cost uses your internal price for key-served requests.

Providers › Activity

The Activity tab lists every key's events: created, tested, rotated, disabled and automatically disabled. It also shows the usage split for your keys.

On this page