One transparent fee. Estimate before you commit.
Provider base rate plus a small service fee. Final cost is based on actual token usage and the internal lanes your batch runs on.
Provider base rate
Live rates from the model catalog.
Service fee
Routing, retries, delivery, and settlement.
Usage estimate
Shows expected cost before work starts.
Direct routes
Choose models directly with SLA, privacy, and price-aware lane selection.
5.90%
service fee on provider subtotal
- Usage-based receipts
- Automatic retries
- Webhook + Slack + email delivery
10% off OpenAI & Anthropic
Customer estimates for native OpenAI and Anthropic batch lanes include an automatic 10% discount on the provider rate before the service fee. The savings show up on every estimate and receipt.
Top up as you go
Top up any amount from $5 to $500 — $10, $25, and $100 are one-tap presets. Credits reserve at micro-dollar precision, so sub-cent jobs cost only what they used.
$0.02 per provider execution lane
Most batches run as one internal lane. Mixed-model JSONL/API jobs keep one customer batch but split execution into model-specific lanes; each dispatched lane carries this small control-plane fee.
SLA tiers: the longer you can wait, the less you pay
Each slower tier is cheaper than the one before it. All six tiers are live: Realtime is a direct chat request, and a batch picks Priority, Standard, Relaxed, Flex or Economy with sla_tier. Each window is the time we promise you from the moment you submit.
Realtime
Best effort
Answered right away at the lane's direct price. No delivery window is promised.
Priority
4h
Same-day pipelines and agent loops.
Standard
24h
The default for batches. The service fee is refunded if we miss the window.
Relaxed
48h
Work that can wait a couple of days.
Flex
72h
Cost-first work needed within days.
Economy
7 days
Large-scale model evaluation, labeling and backfills.
Three ways in
Direct chat completions
POST /v1/chat/completions
The provider's direct price plus the service fee. The maximum cost is held in your wallet first and settled on actual usage.
One request, queued as a batch
POST /v1/chat/completions with "batch": true
The batch price plus the service fee, with no separate quote call.
Bulk batches
POST /v1/batches, or /v1/openai/v1 for OpenAI Batch clients
The batch price plus the service fee, estimated before dispatch.
What the service fee covers
Routing
Provider fit, capacity, privacy, SLA, health, and price.
Execution
Dispatch, polling, retries, fallback, artifacts, and webhooks.
Models
Model catalog, provider onboarding, earnings, and payouts.
On every receipt
Sample - USDPricing questions
See what your batch will cost
Plug in your model, item count, and average tokens - get a cost estimate in seconds.