Execution pricing for AI operators

Simple monthly plans plus usage-based execution units. You stay competitive with auto-routing discounts while heavy models are billed fairly.

Plans Usage Credits FAQ Gift Credits
Base Billing Block
100,000 tokens = 10 units
Auto Route Benefit
Automatic routing receives discounted unit billing
Minimum Charge
1 unit per request

Choose your operator lane

All plans include full API access. Move up when your live volume, routing complexity, or support requirements grow.

Free
$0/mo
Basic
$5/mo
High Volume
Enterprise
$30/mo

How usage billing works

Execution usage is token-based, then adjusted by model tier and routing behavior. Light/cheap routes cost less, heavy/reasoning routes cost more.

  • Base conversion: tokens are converted into usage units using your configured block size.
  • Model multipliers: lightweight, coding, reasoning, and high-parameter models can each have different rates.
  • Auto-route discount: when the system auto-selects an efficient route, discounted usage units are applied.
  • Predictable floor: each request still respects a minimum unit charge for sustainable operations.
1) base = ceil(tokens * units_per_block / tokens_per_block)
2) tiered = ceil(base * multiplier_bps / 10000)
3) auto = ceil(tiered * auto_discount_bps / 10000)
4) billed = max(min_units_per_request, auto)

Need burst capacity?

Top up with credits for campaigns, launches, and temporary demand spikes. Each credit covers one over-limit request and never expires.

100
Credits
$3.00
500
Credits
$10.00

Quick answers

Common operator questions before upgrading.

Do credits expire?

No. Credit packs are persistent and designed for burst traffic, campaign spikes, and launch weeks.

Can I downgrade later?

Yes. You can cancel your active subscription and your account falls back to the free lane.

Why are some requests billed higher?

Large or reasoning-heavy models use higher multipliers because their compute cost is materially higher.

How do I get lower usage cost?

Use auto routing and lighter models for everyday flows, then reserve premium models for critical tasks.