SwiftRelay

← All sprint notes

Claude Sonnet 5 Pricing on SwiftRelay: Line-by-Line Monthly Bill

Cost per lapEN

Claude Sonnet 5 Pricing, Line by Line

Here is a worked monthly bill for Claude Sonnet 5 on SwiftRelay, with every sum shown so you can follow the arithmetic from the first token to the final figure. SwiftRelay is a pay-per-token API relay: there is no monthly fee, nothing to subscribe to, and no bundle you have to buy before you begin. You top up a balance, and each call deducts only what it used.

The three numbers that drive the whole example are the current Claude Sonnet 5 prices shown on the SwiftRelay pricing page: $1.01 per 1M input tokens, $5.06 per 1M output tokens, and $0.0889 per 1M cache-read tokens. Everything below is an example built on those exact rates. Your own bill will follow your own token counts, and the same three lines will show you exactly how it arrived there.

One stopwatch, three clean laps: input, output, cache read. Add them up, and you have the cost of the work.

Building the Worked Monthly Example on SwiftRelay

Let us start with a steady month of Claude Sonnet 5 use. For this example, assume 10,000,000 input tokens, 2,000,000 output tokens and 5,000,000 cache-read tokens.

  • Input line: 10,000,000 input tokens ÷ 1,000,000 = 10 units of 1M tokens.
  • Output line: 2,000,000 output tokens ÷ 1,000,000 = 2 units of 1M tokens.
  • Cache-read line: 5,000,000 cache-read tokens ÷ 1,000,000 = 5 units of 1M tokens.

Now apply the Claude Sonnet 5 rates, one line at a time.

  • Input: 10 × $1.01 = $10.10
  • Output: 2 × $5.06 = $10.12
  • Cache read: 5 × $0.0889 = $0.4445

Next, add the three lines together.

  • Step 1: $10.10 + $10.12 = $20.22
  • Step 2: $20.22 + $0.4445 = $20.6645

The worked total for this example month is $20.6645, which is $20.66 to the nearest cent. That is the entire bill for 10M input tokens, 2M output tokens and 5M cache-read tokens on Claude Sonnet 5 through SwiftRelay. No subscription sits on top of it, because SwiftRelay sells no plans and charges nothing for keeping an account open.

Read the total again and notice what it is made of. The output line is the largest single component here, because output tokens carry the higher rate of $5.06 per 1M. Input sits in the middle at $1.01 per 1M, and cache reads are the smallest at $0.0889 per 1M. When you can see that split, you can see where to steer your own usage if you want the bill to move.

A Lighter Month and a Heavier Month on the Same Rates

Because SwiftRelay bills per token, the bill flexes with your work rather than with a calendar. Here is a second example, a lighter month with half of each figure from the example above: 5,000,000 input tokens, 1,000,000 output tokens and 2,500,000 cache-read tokens.

  • Input: 5 × $1.01 = $5.05
  • Output: 1 × $5.06 = $5.06
  • Cache read: 2.5 × $0.0889 = $0.22225
  • Total: $5.05 + $5.06 = $10.11, then $10.11 + $0.22225 = $10.33225

That lighter example month comes to $10.33225. Now the other direction, a heavier month with double the first example: 20,000,000 input tokens, 4,000,000 output tokens and 10,000,000 cache-read tokens.

  • Input: 20 × $1.01 = $20.20
  • Output: 4 × $5.06 = $20.24
  • Cache read: 10 × $0.0889 = $0.889
  • Total: $20.20 + $20.24 = $40.44, then $40.44 + $0.889 = $41.329

The heavier example month comes to $41.329. Three worked examples, three clean totals, and not one of them requires a plan purchase. This is the good value part of SwiftRelay in plain arithmetic: the spend tracks the tokens you actually sent and received, so a quiet month stays quiet and a busy month pays for the busy work.

Why Per-Token Billing Feels Good Value

Compare the shape of these examples with keeping your own subscription or official account. A fixed subscription asks you to commit the same amount each period, whether you run heavy or light. Per-token billing on SwiftRelay asks for nothing up front beyond the balance you choose to hold, then deducts the cost of each call as it happens.

That makes the $20.6645 example above more than a number. It is a budget you can model before you start, because every line uses published per-1M rates. Multiply your expected input, output and cache-read tokens by $1.01, $5.06 and $0.0889, and you have a working estimate for your own month before you write a single call.

Top-up amounts are available at $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89, so you can size a top-up to the example you just calculated. Sign up and SwiftRelay also credits your account with $0.22 to get the first call moving. If you invite others, referral rewards return 14% of each top-up made by someone you invited and 3% from the people they invite, and those rewards can be withdrawn from a minimum of $4.86 or moved straight into your balance.

There is also a straightforward safety net: unused credit can be refunded within 11 days of a top-up, minus what was used and the payment fee. That keeps the commitment small and the arithmetic honest.

Tracking Tokens, Cost and Keys in the Dashboard

Nothing in this worked bill is a black box. Usage in SwiftRelay lists tokens and cost per call, so you can check the running figure against the worked example line by line and see which calls are moving the total.

The same dashboard holds up to 20 live API keys per account, which is useful when you want to separate projects, a build script and a personal experiment without juggling separate balances. Support is handled through Messages inside the dashboard once you are logged in, so a question about a token count or a charge goes straight to the place where the usage data lives.

For the example month, you could open the usage list after the first week, add up the input, output and cache-read tokens you have already spent, and see whether the pace matches the $20.6645 target. If it does not, you adjust before the month ends, not after. That is the pit-stop advantage of per-token billing: a quick look, a quick correction, and you are back on the lap.

Connecting Claude Sonnet 5 in Three Lines

All of the arithmetic above depends on one thing: getting Claude Sonnet 5 answering through SwiftRelay quickly. Claude models answer only through the official Claude Code or Claude Desktop client, version 2.1.285 or later, and SwiftRelay is connected in minutes by copying three lines and running them.

Here are the Claude Code lines exactly as they should be, each in its own paragraph.

01  export ANTHROPIC_BASE_URL=https://api.fastapirelay.com
02  export ANTHROPIC_AUTH_TOKEN=<your swft_ key>

claude

Paste those three lines into your terminal, replace the placeholder with your own swft_ key from the SwiftRelay dashboard, and run them. The API host is https://api.fastapirelay.com, the base URL points there, and your key authorises the calls. From there, each request is metered at the Claude Sonnet 5 rates used throughout this article: $1.01 per 1M input tokens, $5.06 per 1M output tokens and $0.0889 per 1M cache-read tokens.

Stopwatch ready, laps counted, pit stop finished: sign up for SwiftRelay today, top up your balance and start your first Claude Sonnet 5 call with the cleanest bill in town.

Keep reading

Back for another lap?Log in