What a zero balance means for your next call
SwiftRelay is built around a simple rhythm: top up a balance, make API calls, and let each call deduct what it actually used. There is no recurring fee and nothing to subscribe to. That makes the balance screen the clearest place to watch your spending, especially as it gets close to zero.
When the balance reaches zero, the paid calls that would have been funded by that balance can no longer be covered. The solution is straightforward: top up again and continue. Nothing needs to be repurchased or reorganised, and your usage history remains available so you can review the tokens and cost per call that brought the balance down.
This is particularly useful if you are comparing a pay-per-token relay with keeping a separate subscription or official account for API access. With SwiftRelay, the amount you put in is the amount available, and the amount that leaves is tied to the calls you make. You can also request a refund of unused credit within 11 days of a top-up, minus what was used and the payment fee.
Think of it as a pit stop. You check the gauge, add what you need, and get back on the track. SwiftRelay keeps the stop simple, with balances, usage details, and support through Messages in the dashboard after login. There is no support email to hunt for.
A $6.94 top-up, tracked line by line
The following is a worked example using prices shown on the SwiftRelay pricing page. It starts with a $6.94 top-up and uses DeepSeek V4 Flash and GPT-6.1 Sol. Token counts are illustrative; the prices are the current per 1M token figures.
- Starting balance: $6.94.
- Call 1, DeepSeek V4 Flash: 100,000 input tokens, 20,000 output tokens, and 50,000 cache-read tokens.
- Input: 100,000 divided by 1M times $0.2639 = $0.02639.
- Output: 20,000 divided by 1M times $1.06 = $0.0212.
- Cache read: 50,000 divided by 1M times $0.006944 = $0.0003472.
- Call 1 total: $0.02639 + $0.0212 + $0.0003472 = $0.0479372.
- Balance after Call 1: $6.94 - $0.0479372 = $6.8920628.
That first call takes less than five cents out of the balance, and the usage view lists the tokens and cost per call so you can see where the deduction came from. Now the example continues.
- Call 2, DeepSeek V4 Flash: 200,000 input tokens, 40,000 output tokens, and 100,000 cache-read tokens.
- Input: 200,000 divided by 1M times $0.2639 = $0.05278.
- Output: 40,000 divided by 1M times $1.06 = $0.0424.
- Cache read: 100,000 divided by 1M times $0.006944 = $0.0006944.
- Call 2 total: $0.05278 + $0.0424 + $0.0006944 = $0.0958744.
- Balance after Call 2: $6.8920628 - $0.0958744 = $6.7961884.
Two calls, two clear deductions, and the balance still shows the exact remaining amount. That is the practical value of pay-per-token billing: the figure on the screen is not an estimate for a package you bought, but the amount left after the work you ran.
Running the balance down to its final cents
To show the full line-by-line pattern, the example adds two more calls. Call 3 uses DeepSeek V4 Flash with 500,000 input tokens, 100,000 output tokens, and 200,000 cache-read tokens.
- Input: 500,000 divided by 1M times $0.2639 = $0.13195.
- Output: 100,000 divided by 1M times $1.06 = $0.106.
- Cache read: 200,000 divided by 1M times $0.006944 = $0.0013888.
- Call 3 total: $0.13195 + $0.106 + $0.0013888 = $0.2393388.
- Balance after Call 3: $6.7961884 - $0.2393388 = $6.5568496.
Call 4 switches to GPT-6.1 Sol with 1,000,000 input tokens, 200,000 output tokens, and 500,000 cache-read tokens.
- Input: 1,000,000 divided by 1M times $0.1736 = $0.1736.
- Output: 200,000 divided by 1M times $0.8542 = $0.17084.
- Cache read: 500,000 divided by 1M times $0.0347 = $0.01735.
- Call 4 total: $0.1736 + $0.17084 + $0.01735 = $0.36179.
- Balance after Call 4: $6.5568496 - $0.36179 = $6.1950596.
Now imagine that later usage has reduced the balance to $0.0958744. The next two calls each use the same DeepSeek V4 Flash pattern as Call 1.
- Call 5: $0.02639 + $0.0212 + $0.0003472 = $0.0479372.
- Balance: $0.0958744 - $0.0479372 = $0.0479372.
- Call 6: $0.02639 + $0.0212 + $0.0003472 = $0.0479372.
- Balance: $0.0479372 - $0.0479372 = $0.
There it is: the balance has reached zero, and every step that produced it is visible. At that point, another call needs another top-up. SwiftRelay does not ask you to commit to a package first, so the next move is simply to choose a top-up amount and keep going.
Top up, keep going, and watch every call
SwiftRelay offers top-up amounts of $1.39, $6.94, $13.89, $27.78, $69.44, and $138.89. If you are watching the gauge and know roughly how much runway you want, that choice is right there. A small top-up can cover a short burst of calls, while a larger one leaves more room before the balance needs attention again.
New accounts also receive $0.22 in sign-up gift credit. That is a simple start: connect, copy three lines, run a call, and watch the deduction appear in the usage list alongside the token counts.
Referrals work on the same transparent basis. You receive 14% of each top-up made by someone you invited, and 3% from the people they invite. Rewards can be withdrawn once they reach the $4.86 minimum, or moved into your balance to fund the next call. For an account that is already close to zero, moving rewards into the balance is a quick way to extend the run.
You can run up to 20 live API keys per account, which makes it easy to keep separate keys for separate projects while watching their costs in the same place. The usage view lists tokens and cost per call, so a heavy call and a light call are easy to distinguish.
If something needs clarification, support is available through Messages in the dashboard, where login is required. That keeps the support route in the same place as the balances and usage details you are already checking.
Three lines from sign-up to your next call
Getting connected takes minutes: copy three lines and run. For Claude models, use the official Claude Code or Claude Desktop client, version 2.1.285 or later, with these lines.
01 export ANTHROPIC_BASE_URL=https://api.fastapirelay.com 02 export ANTHROPIC_AUTH_TOKEN=<your swft_ key>
claude
For Codex with GPT, DeepSeek, and other remaining models, use the OpenAI-compatible endpoint with these lines.
01 export OPENAI_BASE_URL=https://api.fastapirelay.com/v1 02 export OPENAI_API_KEY=<your swft_ key>
codex
That is the whole setup path: sign up, top up, copy the three lines, and run. The balance starts visible, each call shows what it used, and a zero balance is handled with another top-up rather than a new commitment.
Stopwatch ready, balance topped up, and your first call waiting on the grid: sign up for SwiftRelay today and take the clean first run.