Skip to content
MAKINA

Pricing

Plans and Prices for Makina's Cloud

Plans for Operator in Makina's cloud, paid monthly through Stripe. Private deployments are priced with you.

Request access

Plans

  • Free

    $0a month

    $1 of use every month. No card.

    Request access
    • Every catalog model, the computer, approvals, takeover, the journal and receipts
    • A daily limit of up to $1

  • Starter

    $25a month

    $15 of use every month.

    Request access
    • Every catalog model, the computer, approvals, takeover, the journal and receipts
    • A daily limit of up to $15

  • Pro

    $100a month

    $60 of use every month.

    Request access
    • Every catalog model, the computer, approvals, takeover, the journal and receipts
    • A daily limit of up to $60

Use is measured in US dollars: the models’ work at the price of the model a task runs on, and computer time while the computer runs. When a period’s use reaches what your plan includes, no new task starts until it resets; a task already running finishes within its budget. Payment is taken by Stripe. Cancel any time from Manage billing on the Usage page: the plan runs to the end of the period you paid for, then returns to Free.

Model prices. Per million tokens, for the models in Makina’s catalog.

GPT-5.6 Luna

Default · OpenAI

The default. Operator uses it unless you choose another.

Input, per million tokens
$0.20
Cached input, per million tokens
$0.02
Output, per million tokens
$1.20
About a step
$0.0016
A short task, measured
$0.005–$0.013

Qwen3.8 Flash

OpenRouter

The lowest price in the catalog.

Input, per million tokens
$0.158
Cached input, per million tokens
$0.017
Output, per million tokens
$0.496
About a step
$0.0012
A short task, measured
$0.006–$0.010

Kimi K3

OpenRouter

The highest output price in the catalog.

Input, per million tokens
$0.707
Cached input, per million tokens
$0.232
Output, per million tokens
$14.77
About a step
$0.0072
A short task, measured
$0.022–$0.058

Grok 4.7

OpenRouter

Input priced like GPT-5.6 Sol, output for less.

Input, per million tokens
$2.11
Cached input, per million tokens
$0.528
Output, per million tokens
$6.33
About a step
$0.016
A short task, measured
$0.041–$0.087

GPT-5.6 Sol

OpenRouter

The highest measured cost per task in the catalog.

Input, per million tokens
$2.11
Cached input, per million tokens
$0.211
Output, per million tokens
$10.55
About a step
$0.016
A short task, measured
$0.055–$0.100

US dollars per million tokens. GPT-5.6 Luna runs on OpenAI’s API at its list price; the other four are served through OpenRouter at its list price plus its 5.5% fee. A step is one look-and-act turn: about 7,000 tokens read (the task so far and a screenshot) and 150 written; long tasks read more. “A short task, measured” is the average cost of two short test tasks run on 9 October 2026. The console shows each catalog model’s price before you choose it, and Usage shows what every task spent.

How Use Is Measured

An hour of computer time while it runs
$0.17
For a paused computer
$0
A task's budget unless you set another, up to $0.25
$0.10
Questions a task may ask you unless you set another limit, up to 64
24

Private Deployments

Priced per engagement: the hardware, bought through Exos Private Cloud or already yours, the deployment and the workflows we build, and the support you want afterwards.

Talk to us

Questions About Pricing

How plans, use and limits work.

Contact
Do I need a card for the Free plan?
No. Every account starts on Free, with $1 of use each month.
What counts as use?
The models' work, at the price of the model a task runs on, and the time your computer is running. A paused computer costs nothing.
What happens when my use runs out?
No new task starts until the period resets. A task already running finishes within its own budget.
Can I set a lower daily limit?
Yes, on the Usage page. Your plan sets the most it can be.
How do I cancel?
From Manage billing on the Usage page. The plan runs to the end of the period you paid for, and the workspace then returns to Free.