Pricing
Start with
one fleet.
256 agents. Four hours on your project. Come back for twenty runs a month when you’re ready.
$1,000 one time
Try a fleet on the project you already have underway.
No subscription. No automatic renewal.
Self-service access coming soon.
$10,000 / month
Run fleets regularly across the projects you’re working on.
20 runs every month.
Self-service access coming soon.
Let’s talk.
For larger workloads or specific requirements.
Run size, volume and terms by agreement.
Discuss your requirements.
Subscribe within 30 days of your first-fleet purchase and we’ll apply the full amount to your first monthly invoice.
Models. Computers.
A complete fleet.
One standard fleet run is 256 agents working for four hours, using the standard model and environment. Each of those agents has a computer, tools and inference, and they can communicate and share files. Your existing agent retrieves the work.
See how a run worksAbout the plans.
Can I change the fleet size or run time?
A standard fleet run is 256 agents for 4 hours. The monthly plan also lets you put the allowance toward different fleet sizes and durations. You’ll see how much of it a configuration uses before launch.
How does the first-fleet discount work?
Subscribe within 30 days of buying your first fleet. Your first month costs $9,000, with all 20 fleet runs. Following months are $10,000.
What if a run fails?
If a technical failure on our side prevents delivery of your first fleet, we’ll provide a replacement run or a refund. The purchase covers fleet time and operation; the result depends on the project.
Do I need a sales call?
No. You’ll get started in Console, connect your existing coding agent and approve the work it prepares. Contact us if you have enterprise requirements.
Which models power the agents?
Our model options include GLM-5.3, MiMo-V2.6-Pro and DeepSeek-V4.1-Flash. Your run’s model and configuration are confirmed before purchase. Have a preference? We can work with you.
Models and toolsWhen can I start?
Self-service checkout is coming. Select a plan to get access updates. You’ll see the supported configuration, available start time and final terms before you pay.
Run sizing and billing details
Prices are in USD. A fleet run uses the standard model and environment configuration. Under the current planning meter, one standard agent-hour uses one credit: 1,024 credits per run and 20,480 for the monthly plan’s 20 runs.
Each agent has its own computer; CPU capacity is shared and allocations can be fractional. The configuration and resource guarantees are shown before purchase. The first fleet is a fixed package in the standard configuration. Credits account for usage of the monthly allowance and approved top-ups. Different model or environment configurations can consume more than one standard run. The estimate is checked before launch.
For one agent per computer, the current resource meter has a minimum of one credit per agent-hour. Reserved CPU at or below one core and guaranteed memory at or below 2 GiB stay within that floor. A 4 GiB memory guarantee uses two credits per agent-hour when the CPU reservation is at or below two cores. Larger reservations can increase the rate. Extra vCPUs and opportunistic CPU or memory burst above the reservation do not add resource charges; burst availability depends on spare capacity.
Monthly subscribers can add credits at $0.50 per credit. Top-ups require your approval. Final metering, credit validity, cancellation and retention terms will be shown before checkout opens.
VM and inference cost calculatorWhat could this do for your project?
Ask in the Claude Code or Codex session you already have open.