Billing

Fleet uses a prepaid credit balance. Charges are deducted as you use compute, storage, and data transfer. You can top up at any time from the dashboard.

What gets charged

Data transfers that happen internally (loading a model or dataset for training, uploading checkpoints, and uploading the final output model) are not charged.

Checking your balance

info = client.billing()
print(info["balance"])              # current credit balance in USD
print(info["storage_gb"])           # total storage in use
print(info["storage_cost_per_month"])  # projected monthly storage cost

fleet billing

Transaction history

result = client.transactions(page=1, limit=50)
for tx in result["transactions"]:
    print(tx["type"], tx["amount"], tx["description"])

fleet billing transactions
fleet billing transactions --page=2 --limit=100

Transaction types: topup, training, inference, storage, network.

Auto-recharge

Auto-recharge automatically tops up your balance when it drops below a threshold. A saved payment card is required. Add one from the dashboard.

Enabling auto-recharge

client.update_settings(
    auto_recharge_enabled=True,
    auto_recharge_threshold=10.0,   # recharge when balance drops below $10
    auto_recharge_amount=25.0,      # top up by $25 each time
)

fleet settings set auto_recharge_enabled true
fleet settings set auto_recharge_threshold 10
fleet settings set auto_recharge_amount 25

The recharge amount must be at least $5 and greater than the threshold.

Status

Check the current auto-recharge status:

settings = client.settings()
print(settings["auto_recharge_status"])

fleet settings

If status is failed, fix the card issue in the dashboard and re-enable:

client.update_settings(auto_recharge_enabled=True)

Re-enabling resets the status from failed to ok.

Storage costs

Storage is metered daily based on what you have stored. Checkpoints are the main source of storage growth. See Storage for how the checkpoint limit and completion pruning keep costs in check.

To see your current storage footprint:

info = client.billing()
print(f"{info['storage_gb']:.2f} GB  —  ${info['storage_cost_per_month']:.4f}/mo")

Deleting a model, dataset, or checkpoint frees that storage immediately.