Pricing
Flows charge usage credits (stored in your organization’s balance) for each operation. Each step in a Flow has a specific cost formula based on the type of work performed.
Below is the pricing logic used for each ChargeType, as defined by our backend:
Table of Charges
| Type | Formula | Notes |
|---|---|---|
| Inference | amount * 0.02 | ”amount” represents the number of images generated or inferences performed. |
| ObjectGenerator | amount * 0.000025 | ”amount” represents total input and output tokens. |
| ObjectTransformer | amount * 0.01 | ”amount” is typically set to 1 per transform, unless you programmatically set it differently. |
| Zipper | (amount * 0.0035) + 0.01 | ”amount” is total output size in MB. There is a base of 0.01 plus 0.0035 per MB. |
| Training | amount * 0.06 | ”amount” = training duration in minutes. So 1 minute of training = $0.06. |
| PrivateFlowRun | amount * 0.01 | For a private Flow run, “amount” is usually 1 per run → 1% overhead. |
| FreePublicFlowRun | amount * 0.28 | For a free public Flow run, each run effectively costs $0.28 to the Flow creator. |
| PaidPublicFlowRun | (amount * 0.23) + 0.39 | For a paid public Flow run, “amount” = flow run price (in USD). 23% of the run price + $0.39 flat fee. |
Examples
1. A Single Inference Step
If your Flow runs one ImageGenerator step that produces 4 images, the cost is:
ChargeType = Inference Usage = 4 inferences Cost = 4 * $0.02 = $0.08
2. Transforming JSON Data
If you have an ObjectTransformer step, by default we set "amount" = 1, meaning:
ChargeType = ObjectTransformer Usage = 1 Cost = 1 * $0.01 = $0.01
(You can override this if you do heavier transformations and track usage more granularly.)
3. Training a Model (2 hours)
Assume your training job runs for 120 minutes (2 hours):
ChargeType = Training Usage = 120 (minutes) Cost = 120 * $0.06 = $7.20
4. Zipping Outputs (50 MB)
If your Flow includes a Zipper step that produces a 50 MB archive:
ChargeType = Zipper Usage = 50 (MB) Cost = (50 * 0.0035) + 0.01 = $0.185 + $0.01 = $0.195
5. Paid Public Flow
Suppose you have a public Flow with a $12 run price. When a user pays $12 to run it:
ChargeType = PaidPublicFlowRun Usage = 12 (the flow run price in USD) Cost = (12 * 0.23) + 0.39 = $2.76 + $0.39 = $3.15
So you, the Flow owner, effectively earn (12 - 3.15) = $8.85 from that run.
(This includes platform & transaction fees, so you see about $8.85 net.)
Balances & Auto-Recharge
Your organization balance is automatically decremented as the steps run. If your balance is insufficient, the Flow stops, and we return an Insufficient Funds error.
Auto-Recharge
You can optionally enable auto-recharge in your organization settings:
- If your balance drops below a threshold, we’ll try to charge a configured amount using your default payment method.
- Any failures are logged, and the system will only retry within a certain time window.
Refunds & Payouts
Public Flow Payouts
If you configure a Paid Public Flow:
- Each time someone pays to run it, a portion goes to you (minus the calculated fees).
- Payout is automatically handled by Stripe Connect if you’ve set up a connected account.
- If no connected account is available, the earnings go to your organization’s wallet balance.
Refunds
Payment refunds (e.g., if the user cancels or errors occur before the Flow runs) are initiated through our Stripe integration. Once processed, the corresponding charge is marked canceled, and your organization balance is adjusted if you had already received funds.
Summary
- Each step in a Flow has a cost determined by its ChargeType multiplied by a usage count.
- Public Flows charge external users. The flow owner is paid out after a platform & transaction fee.
- Private Flows simply deduct from your organization’s balance.
- Auto-Recharge helps ensure your balance never runs dry.
With these principles in mind, you can accurately estimate how each Flow run impacts your credits or potential earnings. Check the Billing & Usage section for deeper coverage on how to manage your credits, set up payment methods, or integrate Stripe Connect.